+include $(SUBDIR)../config.mak
+
NAME = avcodec
FFLIBS = avutil
options.o \
parser.o \
raw.o \
+ rawdec.o \
resample.o \
resample2.o \
simple_idct.o \
# parts needed for many different codecs
OBJS-$(CONFIG_AANDCT) += aandcttab.o
OBJS-$(CONFIG_AC3DSP) += ac3dsp.o
+OBJS-$(CONFIG_CRYSTALHD) += crystalhd.o
OBJS-$(CONFIG_ENCODERS) += faandct.o jfdctfst.o jfdctint.o
OBJS-$(CONFIG_DCT) += dct.o dct32_fixed.o dct32_float.o
OBJS-$(CONFIG_DWT) += dwt.o
OBJS-$(CONFIG_ANM_DECODER) += anm.o
OBJS-$(CONFIG_ANSI_DECODER) += ansi.o cga_data.o
OBJS-$(CONFIG_APE_DECODER) += apedec.o
-OBJS-$(CONFIG_ASS_DECODER) += assdec.o ass.o
+OBJS-$(CONFIG_ASS_DECODER) += assdec.o ass.o ass_split.o
OBJS-$(CONFIG_ASS_ENCODER) += assenc.o ass.o
OBJS-$(CONFIG_ASV1_DECODER) += asv1.o mpeg12data.o
OBJS-$(CONFIG_ASV1_ENCODER) += asv1.o mpeg12data.o
OBJS-$(CONFIG_BINK_DECODER) += bink.o binkdsp.o
OBJS-$(CONFIG_BINKAUDIO_DCT_DECODER) += binkaudio.o wma.o
OBJS-$(CONFIG_BINKAUDIO_RDFT_DECODER) += binkaudio.o wma.o
+OBJS-$(CONFIG_BINTEXT_DECODER) += bintext.o cga_data.o
OBJS-$(CONFIG_BMP_DECODER) += bmp.o msrledec.o
OBJS-$(CONFIG_BMP_ENCODER) += bmpenc.o
OBJS-$(CONFIG_BMV_VIDEO_DECODER) += bmv.o
OBJS-$(CONFIG_CSCD_DECODER) += cscd.o
OBJS-$(CONFIG_CYUV_DECODER) += cyuv.o
OBJS-$(CONFIG_DCA_DECODER) += dca.o synth_filter.o dcadsp.o
+OBJS-$(CONFIG_DCA_ENCODER) += dcaenc.o
+OBJS-$(CONFIG_DIRAC_DECODER) += diracdec.o dirac.o diracdsp.o \
+ dirac_arith.o mpeg12data.o dwt.o
OBJS-$(CONFIG_DFA_DECODER) += dfa.o
OBJS-$(CONFIG_DNXHD_DECODER) += dnxhddec.o dnxhddata.o
OBJS-$(CONFIG_DNXHD_ENCODER) += dnxhdenc.o dnxhddata.o \
OBJS-$(CONFIG_EIGHTBPS_DECODER) += 8bps.o
OBJS-$(CONFIG_EIGHTSVX_EXP_DECODER) += 8svx.o
OBJS-$(CONFIG_EIGHTSVX_FIB_DECODER) += 8svx.o
+OBJS-$(CONFIG_EIGHTSVX_RAW_DECODER) += 8svx.o
OBJS-$(CONFIG_ESCAPE124_DECODER) += escape124.o
OBJS-$(CONFIG_FFV1_DECODER) += ffv1.o rangecoder.o
OBJS-$(CONFIG_FFV1_ENCODER) += ffv1.o rangecoder.o
OBJS-$(CONFIG_FFVHUFF_DECODER) += huffyuv.o
OBJS-$(CONFIG_FFVHUFF_ENCODER) += huffyuv.o
-OBJS-$(CONFIG_FLAC_DECODER) += flacdec.o flacdata.o flac.o
-OBJS-$(CONFIG_FLAC_ENCODER) += flacenc.o flacdata.o flac.o
+OBJS-$(CONFIG_FLAC_DECODER) += flacdec.o flacdata.o flac.o vorbis_data.o
+OBJS-$(CONFIG_FLAC_ENCODER) += flacenc.o flacdata.o flac.o vorbis_data.o
OBJS-$(CONFIG_FLASHSV_DECODER) += flashsv.o
OBJS-$(CONFIG_FLASHSV_ENCODER) += flashsvenc.o
+OBJS-$(CONFIG_FLASHSV2_ENCODER) += flashsv2enc.o
OBJS-$(CONFIG_FLASHSV2_DECODER) += flashsv.o
OBJS-$(CONFIG_FLIC_DECODER) += flicvideo.o
OBJS-$(CONFIG_FOURXM_DECODER) += 4xm.o
OBJS-$(CONFIG_FRAPS_DECODER) += fraps.o
OBJS-$(CONFIG_FRWU_DECODER) += frwu.o
+OBJS-$(CONFIG_G723_1_DECODER) += g723_1.o acelp_vectors.o \
+ celp_filters.o celp_math.o
+OBJS-$(CONFIG_G723_1_ENCODER) += g723_1.o
+OBJS-$(CONFIG_G729_DECODER) += g729dec.o lsp.o celp_math.o acelp_filters.o acelp_pitch_delay.o acelp_vectors.o g729postfilter.o
OBJS-$(CONFIG_GIF_DECODER) += gifdec.o lzw.o
OBJS-$(CONFIG_GIF_ENCODER) += gif.o lzwenc.o
OBJS-$(CONFIG_GSM_DECODER) += gsmdec.o gsmdec_data.o msgsmdec.o
OBJS-$(CONFIG_HUFFYUV_DECODER) += huffyuv.o
OBJS-$(CONFIG_HUFFYUV_ENCODER) += huffyuv.o
OBJS-$(CONFIG_IDCIN_DECODER) += idcinvideo.o
+OBJS-$(CONFIG_IDF_DECODER) += bintext.o cga_data.o
OBJS-$(CONFIG_IFF_BYTERUN1_DECODER) += iff.o
OBJS-$(CONFIG_IFF_ILBM_DECODER) += iff.o
OBJS-$(CONFIG_IMC_DECODER) += imc.o
OBJS-$(CONFIG_INDEO5_DECODER) += indeo5.o ivi_common.o ivi_dsp.o
OBJS-$(CONFIG_INTERPLAY_DPCM_DECODER) += dpcm.o
OBJS-$(CONFIG_INTERPLAY_VIDEO_DECODER) += interplayvideo.o
+OBJS-$(CONFIG_JPEG2000_DECODER) += j2kdec.o mqcdec.o mqc.o j2k.o j2k_dwt.o
+OBJS-$(CONFIG_JPEG2000_ENCODER) += j2kenc.o mqcenc.o mqc.o j2k.o j2k_dwt.o
OBJS-$(CONFIG_JPEGLS_DECODER) += jpeglsdec.o jpegls.o \
mjpegdec.o mjpeg.o
OBJS-$(CONFIG_JPEGLS_ENCODER) += jpeglsenc.o jpegls.o
OBJS-$(CONFIG_MPC8_DECODER) += mpc8.o mpc.o mpegaudiodec.o \
mpegaudiodecheader.o mpegaudio.o \
mpegaudiodata.o
+OBJS-$(CONFIG_MPEGVIDEO_DECODER) += mpeg12.o mpeg12data.o \
+ mpegvideo.o error_resilience.o
OBJS-$(CONFIG_MPEG_XVMC_DECODER) += mpegvideo_xvmc.o
OBJS-$(CONFIG_MPEG1VIDEO_DECODER) += mpeg12.o mpeg12data.o \
mpegvideo.o error_resilience.o
OBJS-$(CONFIG_MPEG1VIDEO_ENCODER) += mpeg12enc.o mpegvideo_enc.o \
+ timecode.o \
motion_est.o ratecontrol.o \
mpeg12.o mpeg12data.o \
mpegvideo.o error_resilience.o
OBJS-$(CONFIG_MPEG2VIDEO_DECODER) += mpeg12.o mpeg12data.o \
mpegvideo.o error_resilience.o
OBJS-$(CONFIG_MPEG2VIDEO_ENCODER) += mpeg12enc.o mpegvideo_enc.o \
+ timecode.o \
motion_est.o ratecontrol.o \
mpeg12.o mpeg12data.o \
mpegvideo.o error_resilience.o
h263.o ituh263dec.o mpeg4videodec.o
OBJS-$(CONFIG_MSRLE_DECODER) += msrle.o msrledec.o
OBJS-$(CONFIG_MSVIDEO1_DECODER) += msvideo1.o
+OBJS-$(CONFIG_MSVIDEO1_ENCODER) += msvideo1enc.o elbg.o
OBJS-$(CONFIG_MSZH_DECODER) += lcldec.o
OBJS-$(CONFIG_MXPEG_DECODER) += mxpegdec.o mjpegdec.o mjpeg.o
OBJS-$(CONFIG_NELLYMOSER_DECODER) += nellymoserdec.o nellymoser.o
OBJS-$(CONFIG_PNG_ENCODER) += png.o pngenc.o
OBJS-$(CONFIG_PPM_DECODER) += pnmdec.o pnm.o
OBJS-$(CONFIG_PPM_ENCODER) += pnmenc.o pnm.o
-OBJS-$(CONFIG_PRORES_DECODER) += proresdec.o proresdsp.o
+OBJS-$(CONFIG_PRORES_DECODER) += proresdec2.o
+OBJS-$(CONFIG_PRORES_LGPL_DECODER) += proresdec_lgpl.o proresdsp.o
+OBJS-$(CONFIG_PRORES_ENCODER) += proresenc.o
OBJS-$(CONFIG_PTX_DECODER) += ptx.o
OBJS-$(CONFIG_QCELP_DECODER) += qcelpdec.o celp_math.o \
celp_filters.o acelp_vectors.o \
OBJS-$(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-$(CONFIG_SVQ1_DECODER) += svq1dec.o svq1.o h263.o \
mpegvideo.o error_resilience.o
OBJS-$(CONFIG_VP8_DECODER) += vp8.o vp8dsp.o vp56rac.o
OBJS-$(CONFIG_VQA_DECODER) += vqavideo.o
OBJS-$(CONFIG_WAVPACK_DECODER) += wavpack.o
+OBJS-$(CONFIG_WMALOSSLESS_DECODER) += wmalosslessdec.o wma.o
OBJS-$(CONFIG_WMAPRO_DECODER) += wmaprodec.o wma.o
OBJS-$(CONFIG_WMAV1_DECODER) += wmadec.o wma.o aactab.o
OBJS-$(CONFIG_WMAV1_ENCODER) += wmaenc.o wma.o aactab.o
OBJS-$(CONFIG_XAN_DPCM_DECODER) += dpcm.o
OBJS-$(CONFIG_XAN_WC3_DECODER) += xan.o
OBJS-$(CONFIG_XAN_WC4_DECODER) += xxan.o
+OBJS-$(CONFIG_XBIN_DECODER) += bintext.o cga_data.o
OBJS-$(CONFIG_XL_DECODER) += xl.o
OBJS-$(CONFIG_XSUB_DECODER) += xsubdec.o
OBJS-$(CONFIG_XSUB_ENCODER) += xsubenc.o
OBJS-$(CONFIG_ADTS_MUXER) += mpeg4audio.o
OBJS-$(CONFIG_CAF_DEMUXER) += mpeg4audio.o mpegaudiodata.o
OBJS-$(CONFIG_DV_DEMUXER) += dvdata.o
-OBJS-$(CONFIG_DV_MUXER) += dvdata.o
-OBJS-$(CONFIG_FLAC_DEMUXER) += flacdec.o flacdata.o flac.o
-OBJS-$(CONFIG_FLAC_MUXER) += flacdec.o flacdata.o flac.o
+OBJS-$(CONFIG_DV_MUXER) += dvdata.o timecode.o
+OBJS-$(CONFIG_FLAC_DEMUXER) += flacdec.o flacdata.o flac.o vorbis_data.o
+OBJS-$(CONFIG_FLAC_MUXER) += flacdec.o flacdata.o flac.o vorbis_data.o
OBJS-$(CONFIG_FLV_DEMUXER) += mpeg4audio.o
OBJS-$(CONFIG_GXF_DEMUXER) += mpeg12data.o
OBJS-$(CONFIG_IFF_DEMUXER) += iff.o
OBJS-$(CONFIG_LATM_MUXER) += mpeg4audio.o
-OBJS-$(CONFIG_MATROSKA_AUDIO_MUXER) += xiph.o mpeg4audio.o \
+OBJS-$(CONFIG_MATROSKA_AUDIO_MUXER) += xiph.o mpeg4audio.o vorbis_data.o \
flacdec.o flacdata.o flac.o
OBJS-$(CONFIG_MATROSKA_DEMUXER) += mpeg4audio.o mpegaudiodata.o
OBJS-$(CONFIG_MATROSKA_MUXER) += xiph.o mpeg4audio.o \
flacdec.o flacdata.o flac.o \
- mpegaudiodata.o
+ mpegaudiodata.o vorbis_data.o
OBJS-$(CONFIG_MP3_MUXER) += mpegaudiodata.o mpegaudiodecheader.o
OBJS-$(CONFIG_MOV_DEMUXER) += mpeg4audio.o mpegaudiodata.o
OBJS-$(CONFIG_MOV_MUXER) += mpeg4audio.o mpegaudiodata.o
OBJS-$(CONFIG_MPEGTS_MUXER) += mpegvideo.o mpeg4audio.o
OBJS-$(CONFIG_MPEGTS_DEMUXER) += mpeg4audio.o mpegaudiodata.o
+OBJS-$(CONFIG_MXF_MUXER) += timecode.o
OBJS-$(CONFIG_NUT_MUXER) += mpegaudiodata.o
OBJS-$(CONFIG_OGG_DEMUXER) += flacdec.o flacdata.o flac.o \
- dirac.o mpeg12data.o
-OBJS-$(CONFIG_OGG_MUXER) += xiph.o flacdec.o flacdata.o flac.o
+ dirac.o mpeg12data.o vorbis_data.o
+OBJS-$(CONFIG_OGG_MUXER) += xiph.o flacdec.o flacdata.o flac.o \
+ vorbis_data.o
OBJS-$(CONFIG_RTP_MUXER) += mpeg4audio.o mpegvideo.o xiph.o
OBJS-$(CONFIG_SPDIF_DEMUXER) += aacadtsdec.o mpeg4audio.o
OBJS-$(CONFIG_WEBM_MUXER) += xiph.o mpeg4audio.o \
flacdec.o flacdata.o flac.o \
- mpegaudiodata.o
+ mpegaudiodata.o vorbis_data.o
OBJS-$(CONFIG_WTV_DEMUXER) += mpeg4audio.o mpegaudiodata.o
# external codec libraries
+OBJS-$(CONFIG_LIBAACPLUS_ENCODER) += libaacplus.o
+OBJS-$(CONFIG_LIBCELT_DECODER) += libcelt_dec.o
OBJS-$(CONFIG_LIBDIRAC_DECODER) += libdiracdec.o
OBJS-$(CONFIG_LIBDIRAC_ENCODER) += libdiracenc.o libdirac_libschro.o
OBJS-$(CONFIG_LIBFAAC_ENCODER) += libfaac.o
OBJS-$(CONFIG_LIBOPENCORE_AMRNB_DECODER) += libopencore-amr.o
OBJS-$(CONFIG_LIBOPENCORE_AMRNB_ENCODER) += libopencore-amr.o
OBJS-$(CONFIG_LIBOPENCORE_AMRWB_DECODER) += libopencore-amr.o
-OBJS-$(CONFIG_LIBOPENJPEG_DECODER) += libopenjpeg.o
+OBJS-$(CONFIG_LIBOPENJPEG_DECODER) += libopenjpegdec.o
+OBJS-$(CONFIG_LIBOPENJPEG_ENCODER) += libopenjpegenc.o
OBJS-$(CONFIG_LIBSCHROEDINGER_DECODER) += libschroedingerdec.o \
libschroedinger.o \
libdirac_libschro.o
libdirac_libschro.o
OBJS-$(CONFIG_LIBSPEEX_DECODER) += libspeexdec.o
OBJS-$(CONFIG_LIBSPEEX_ENCODER) += libspeexenc.o
+OBJS-$(CONFIG_LIBSTAGEFRIGHT_H264_DECODER)+= libstagefright.o
OBJS-$(CONFIG_LIBTHEORA_ENCODER) += libtheoraenc.o
+OBJS-$(CONFIG_LIBUTVIDEO_DECODER) += libutvideo.o
OBJS-$(CONFIG_LIBVO_AACENC_ENCODER) += libvo-aacenc.o mpeg4audio.o
OBJS-$(CONFIG_LIBVO_AMRWBENC_ENCODER) += libvo-amrwbenc.o
OBJS-$(CONFIG_LIBVORBIS_ENCODER) += libvorbis.o vorbis_data.o
OBJS-$(CONFIG_DNXHD_PARSER) += dnxhd_parser.o
OBJS-$(CONFIG_DVBSUB_PARSER) += dvbsub_parser.o
OBJS-$(CONFIG_DVDSUB_PARSER) += dvdsub_parser.o
-OBJS-$(CONFIG_FLAC_PARSER) += flac_parser.o flacdata.o flac.o
+OBJS-$(CONFIG_FLAC_PARSER) += flac_parser.o flacdata.o flac.o \
+ vorbis_data.o
OBJS-$(CONFIG_H261_PARSER) += h261_parser.o
OBJS-$(CONFIG_H263_PARSER) += h263_parser.o
OBJS-$(CONFIG_H264_PARSER) += h264_parser.o h264.o \
# thread libraries
OBJS-$(HAVE_PTHREADS) += pthread.o
OBJS-$(HAVE_W32THREADS) += pthread.o
+OBJS-$(HAVE_OS2THREADS) += pthread.o
OBJS-$(CONFIG_MLIB) += mlib/dsputil_mlib.o \
SKIPHEADERS-$(CONFIG_LIBDIRAC) += libdirac.h
SKIPHEADERS-$(CONFIG_LIBSCHROEDINGER) += libschroedinger.h
SKIPHEADERS-$(CONFIG_VAAPI) += vaapi_internal.h
-SKIPHEADERS-$(CONFIG_VDA) += vda.h vda_internal.h
+SKIPHEADERS-$(CONFIG_VDA) += vda_internal.h
SKIPHEADERS-$(CONFIG_VDPAU) += vdpau.h
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
$(SUBDIR)pcm.o: $(SUBDIR)pcm_tables.h
$(SUBDIR)qdm2.o: $(SUBDIR)qdm2_tables.h
endif
+
+CODEC_NAMES_SH := $(SRC_PATH)/$(SUBDIR)codec_names.sh
+AVCODEC_H := $(SRC_PATH)/$(SUBDIR)avcodec.h
+$(SUBDIR)codec_names.h: $(CODEC_NAMES_SH) config.h $(AVCODEC_H)
+ $(CC) $(CPPFLAGS) $(CFLAGS) -E $(AVCODEC_H) | \
+ $(CODEC_NAMES_SH) config.h $@
+$(SUBDIR)utils.o: $(SUBDIR)codec_names.h
* H.26L/H.264/AVC/JVT/14496-10/... decoder
* Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * Libav is distributed in the hope that it will be useful,
+ * FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
*/
#include "libavutil/imgutils.h"
+#include "libavutil/opt.h"
#include "internal.h"
#include "dsputil.h"
#include "avcodec.h"
}
return 0;
-} //FIXME cleanup like ff_h264_check_intra_pred_mode
+} //FIXME cleanup like check_intra_pred_mode
-/**
- * checks if the top & left blocks are available if needed & changes the dc mode so it only uses the available blocks.
- */
-int ff_h264_check_intra_pred_mode(H264Context *h, int mode){
+static int check_intra_pred_mode(H264Context *h, int mode, int is_chroma){
MpegEncContext * const s = &h->s;
static const int8_t top [7]= {LEFT_DC_PRED8x8, 1,-1,-1};
static const int8_t left[7]= { TOP_DC_PRED8x8,-1, 2,-1,DC_128_PRED8x8};
if((h->left_samples_available&0x8080) != 0x8080){
mode= left[ mode ];
- if(h->left_samples_available&0x8080){ //mad cow disease mode, aka MBAFF + constrained_intra_pred
+ if(is_chroma && (h->left_samples_available&0x8080)){ //mad cow disease mode, aka MBAFF + constrained_intra_pred
mode= ALZHEIMER_DC_L0T_PRED8x8 + (!(h->left_samples_available&0x8000)) + 2*(mode == DC_128_PRED8x8);
}
if(mode<0){
return mode;
}
+/**
+ * checks if the top & left blocks are available if needed & changes the dc mode so it only uses the available blocks.
+ */
+int ff_h264_check_intra16x16_pred_mode(H264Context *h, int mode)
+{
+ return check_intra_pred_mode(h, mode, 0);
+}
+
+/**
+ * checks if the top & left blocks are available if needed & changes the dc mode so it only uses the available blocks.
+ */
+int ff_h264_check_intra_chroma_pred_mode(H264Context *h, int mode)
+{
+ return check_intra_pred_mode(h, mode, 1);
+}
+
+
const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src, int *dst_length, int *consumed, int length){
int i, si, di;
uint8_t *dst;
i-= RS;
}
- if(i>=length-1){ //no escaped 0
- *dst_length= length;
- *consumed= length+1; //+1 for the header
- return src;
- }
-
bufidx = h->nal_unit_type == NAL_DPC ? 1 : 0; // use second escape buffer for inter data
- av_fast_malloc(&h->rbsp_buffer[bufidx], &h->rbsp_buffer_size[bufidx], length+FF_INPUT_BUFFER_PADDING_SIZE);
+ si=h->rbsp_buffer_size[bufidx];
+ av_fast_malloc(&h->rbsp_buffer[bufidx], &h->rbsp_buffer_size[bufidx], length+FF_INPUT_BUFFER_PADDING_SIZE+MAX_MBPAIR_SIZE);
dst= h->rbsp_buffer[bufidx];
+ if(si != h->rbsp_buffer_size[bufidx])
+ memset(dst + length, 0, FF_INPUT_BUFFER_PADDING_SIZE+MAX_MBPAIR_SIZE);
if (dst == NULL){
return NULL;
}
+ if(i>=length-1){ //no escaped 0
+ *dst_length= length;
+ *consumed= length+1; //+1 for the header
+ if(h->s.avctx->flags2 & CODEC_FLAG2_FAST){
+ return src;
+ }else{
+ memcpy(dst, src, length);
+ return dst;
+ }
+ }
+
//printf("decoding esc\n");
memcpy(dst, src, i);
si=di=i;
s->dsp.prefetch(src[1]+off, s->linesize, 4);
s->dsp.prefetch(src[2]+off, s->linesize, 4);
}else{
- off= ((mx>>1) << pixel_shift) + ((my>>1) + (s->mb_x&7))*s->uvlinesize + (64 << pixel_shift);
+ off= (((mx>>1)+64)<<pixel_shift) + ((my>>1) + (s->mb_x&7))*s->uvlinesize;
s->dsp.prefetch(src[1]+off, src[2]-src[1], 2);
}
}
s->height = s->avctx->height;
s->codec_id= s->avctx->codec->id;
- ff_h264dsp_init(&h->h264dsp, 8, 1);
- ff_h264_pred_init(&h->hpc, s->codec_id, 8, 1);
+ s->avctx->bits_per_raw_sample = 8;
+ h->cur_chroma_format_idc = 1;
+
+ ff_h264dsp_init(&h->h264dsp,
+ s->avctx->bits_per_raw_sample, h->cur_chroma_format_idc);
+ ff_h264_pred_init(&h->hpc, s->codec_id,
+ s->avctx->bits_per_raw_sample, h->cur_chroma_format_idc);
h->dequant_coeff_pps= -1;
s->unrestricted_mv=1;
+ s->dsp.dct_bits = 16;
dsputil_init(&s->dsp, s->avctx); // needed so that idct permutation is known early
memset(h->pps.scaling_matrix4, 16, 6*16*sizeof(uint8_t));
memset(h->pps.scaling_matrix8, 16, 2*64*sizeof(uint8_t));
}
-int ff_h264_decode_extradata(H264Context *h)
+int ff_h264_decode_extradata(H264Context *h, const uint8_t *buf, int size)
{
AVCodecContext *avctx = h->s.avctx;
- if(avctx->extradata[0] == 1){
+ if(!buf || size <= 0)
+ return -1;
+
+ if(buf[0] == 1){
int i, cnt, nalsize;
- unsigned char *p = avctx->extradata;
+ const unsigned char *p = buf;
h->is_avc = 1;
- if(avctx->extradata_size < 7) {
+ if(size < 7) {
av_log(avctx, AV_LOG_ERROR, "avcC too short\n");
return -1;
}
p += 6;
for (i = 0; i < cnt; i++) {
nalsize = AV_RB16(p) + 2;
- if (p - avctx->extradata + nalsize > avctx->extradata_size)
+ if(nalsize > size - (p-buf))
return -1;
if(decode_nal_units(h, p, nalsize) < 0) {
av_log(avctx, AV_LOG_ERROR, "Decoding sps %d from avcC failed\n", i);
cnt = *(p++); // Number of pps
for (i = 0; i < cnt; i++) {
nalsize = AV_RB16(p) + 2;
- if (p - avctx->extradata + nalsize > avctx->extradata_size)
+ if(nalsize > size - (p-buf))
return -1;
if (decode_nal_units(h, p, nalsize) < 0) {
av_log(avctx, AV_LOG_ERROR, "Decoding pps %d from avcC failed\n", i);
p += nalsize;
}
// Now store right nal length size, that will be use to parse all other nals
- h->nal_length_size = (avctx->extradata[4] & 0x03) + 1;
+ h->nal_length_size = (buf[4] & 0x03) + 1;
} else {
h->is_avc = 0;
- if(decode_nal_units(h, avctx->extradata, avctx->extradata_size) < 0)
+ if(decode_nal_units(h, buf, size) < 0)
return -1;
}
return 0;
}
if(avctx->extradata_size > 0 && avctx->extradata &&
- ff_h264_decode_extradata(h))
+ ff_h264_decode_extradata(h, avctx->extradata, avctx->extradata_size))
return -1;
if(h->sps.bitstream_restriction_flag && s->avctx->has_b_frames < h->sps.num_reorder_frames){
copy_picture_range(h->delayed_pic, h1->delayed_pic, MAX_DELAYED_PIC_COUNT+2, s, s1);
h->last_slice_type = h1->last_slice_type;
+ h->sync = h1->sync;
if(!s->current_picture_ptr) return 0;
s->low_delay= 0;
}
+ for (i = 0; 1; i++) {
+ if(i == MAX_DELAYED_PIC_COUNT || cur->poc < h->last_pocs[i]){
+ if(i)
+ h->last_pocs[i-1] = cur->poc;
+ break;
+ } else if(i) {
+ h->last_pocs[i-1]= h->last_pocs[i];
+ }
+ }
+ out_of_order = MAX_DELAYED_PIC_COUNT - i;
+ if( cur->f.pict_type == AV_PICTURE_TYPE_B
+ || (h->last_pocs[MAX_DELAYED_PIC_COUNT-2] > INT_MIN && h->last_pocs[MAX_DELAYED_PIC_COUNT-1] - h->last_pocs[MAX_DELAYED_PIC_COUNT-2] > 2))
+ out_of_order = FFMAX(out_of_order, 1);
+ if(s->avctx->has_b_frames < out_of_order && !h->sps.bitstream_restriction_flag){
+ av_log(s->avctx, AV_LOG_WARNING, "Increasing reorder buffer to %d\n", out_of_order);
+ s->avctx->has_b_frames = out_of_order;
+ s->low_delay = 0;
+ }
+
pics = 0;
while(h->delayed_pic[pics]) pics++;
- assert(pics <= MAX_DELAYED_PIC_COUNT);
+ av_assert0(pics <= MAX_DELAYED_PIC_COUNT);
h->delayed_pic[pics++] = cur;
if (cur->f.reference == 0)
h->next_outputed_poc= INT_MIN;
out_of_order = out->poc < h->next_outputed_poc;
- if(h->sps.bitstream_restriction_flag && s->avctx->has_b_frames >= h->sps.num_reorder_frames)
- { }
- else if (out_of_order && pics-1 == s->avctx->has_b_frames &&
- s->avctx->has_b_frames < MAX_DELAYED_PIC_COUNT) {
- int cnt = 0, invalid = 0;
- for (i = 0; i < MAX_DELAYED_PIC_COUNT; i++) {
- cnt += out->poc < h->last_pocs[i];
- invalid += h->last_pocs[i] == INT_MIN;
- }
- if (invalid + cnt < MAX_DELAYED_PIC_COUNT) {
- s->avctx->has_b_frames = FFMAX(s->avctx->has_b_frames, cnt);
- } else if (cnt) {
- for (i = 0; i < MAX_DELAYED_PIC_COUNT; i++)
- h->last_pocs[i] = INT_MIN;
- }
- s->low_delay = 0;
- } else if (s->low_delay &&
- ((h->next_outputed_poc != INT_MIN && out->poc > h->next_outputed_poc + 2) ||
- cur->f.pict_type == AV_PICTURE_TYPE_B)) {
- s->low_delay = 0;
- s->avctx->has_b_frames++;
- }
-
if(out_of_order || pics > s->avctx->has_b_frames){
out->f.reference &= ~DELAYED_PIC_REF;
out->owner2 = s; // for frame threading, the owner must be the second field's thread
for(i=out_idx; h->delayed_pic[i]; i++)
h->delayed_pic[i] = h->delayed_pic[i+1];
}
- memmove(h->last_pocs, &h->last_pocs[1], sizeof(*h->last_pocs) * (MAX_DELAYED_PIC_COUNT - 1));
- h->last_pocs[MAX_DELAYED_PIC_COUNT - 1] = out->poc;
if(!out_of_order && pics > s->avctx->has_b_frames){
h->next_output_pic = out;
if (out_idx == 0 && h->delayed_pic[0] && (h->delayed_pic[0]->f.key_frame || h->delayed_pic[0]->mmco_reset)) {
av_log(s->avctx, AV_LOG_DEBUG, "no picture\n");
}
+ if (h->next_output_pic && h->next_output_pic->sync) {
+ h->sync |= 2;
+ }
+
if (setup_finished)
ff_thread_finish_setup(s->avctx);
}
AV_COPY128(top_border+16, src_cb + 15*uvlinesize);
AV_COPY128(top_border+32, src_cr + 15*uvlinesize);
}
- } else if(chroma422) {
+ } else if(chroma422){
if (pixel_shift) {
AV_COPY128(top_border+32, src_cb + 15*uvlinesize);
AV_COPY128(top_border+48, src_cr + 15*uvlinesize);
}
if (!simple && IS_INTRA_PCM(mb_type)) {
+ const int bit_depth = h->sps.bit_depth_luma;
if (pixel_shift) {
- const int bit_depth = h->sps.bit_depth_luma;
int j;
GetBitContext gb;
init_get_bits(&gb, (uint8_t*)h->mb, 384*bit_depth);
if (!h->sps.chroma_format_idc) {
for (i = 0; i < block_h; i++) {
uint16_t *tmp_cb = (uint16_t*)(dest_cb + i*uvlinesize);
- for (j = 0; j < 8; j++) {
- tmp_cb[j] = 1 << (bit_depth - 1);
- }
- }
- for (i = 0; i < block_h; i++) {
uint16_t *tmp_cr = (uint16_t*)(dest_cr + i*uvlinesize);
for (j = 0; j < 8; j++) {
- tmp_cr[j] = 1 << (bit_depth - 1);
+ tmp_cb[j] = tmp_cr[j] = 1 << (bit_depth - 1);
}
}
} else {
}
if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
if (!h->sps.chroma_format_idc) {
- for (i = 0; i < block_h; i++) {
- memset(dest_cb + i*uvlinesize, 128, 8);
- memset(dest_cr + i*uvlinesize, 128, 8);
+ for (i=0; i<8; i++) {
+ memset(dest_cb + i*uvlinesize, 1 << (bit_depth - 1), 8);
+ memset(dest_cr + i*uvlinesize, 1 << (bit_depth - 1), 8);
}
} else {
- for (i = 0; i < block_h; i++) {
+ for (i=0; i<block_h; i++) {
memcpy(dest_cb + i*uvlinesize, h->mb + 128 + i*4, 8);
memcpy(dest_cr + i*uvlinesize, h->mb + 160 + i*4, 8);
}
h->h264dsp.h264_idct_add8(dest, block_offset,
h->mb, uvlinesize,
h->non_zero_count_cache);
- }else{
+ }
+#if CONFIG_SVQ3_DECODER
+ else{
h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16*16*1, h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][h->chroma_qp[0]][0]);
h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16*16*2, h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][h->chroma_qp[1]][0]);
for(j=1; j<3; j++){
}
}
}
+#endif
}
}
}
* instantaneous decoder refresh.
*/
static void idr(H264Context *h){
+ int i;
ff_h264_remove_all_refs(h);
- h->prev_frame_num= 0;
+ h->prev_frame_num= -1;
h->prev_frame_num_offset= 0;
h->prev_poc_msb=
h->prev_poc_lsb= 0;
+ for (i = 0; i < MAX_DELAYED_PIC_COUNT; i++)
+ h->last_pocs[i] = INT_MIN;
}
/* forget old pics after a seek */
static void flush_dpb(AVCodecContext *avctx){
H264Context *h= avctx->priv_data;
int i;
- for(i=0; i<MAX_DELAYED_PIC_COUNT; i++) {
+ for(i=0; i<=MAX_DELAYED_PIC_COUNT; i++) {
if(h->delayed_pic[i])
h->delayed_pic[i]->f.reference = 0;
h->delayed_pic[i]= NULL;
h->s.first_field= 0;
ff_h264_reset_sei(h);
ff_mpeg_flush(avctx);
+ h->recovery_frame= -1;
+ h->sync= 0;
}
static int init_poc(H264Context *h){
s->me.qpel_avg= s->dsp.avg_h264_qpel_pixels_tab;
}
- first_mb_in_slice= get_ue_golomb(&s->gb);
+ first_mb_in_slice= get_ue_golomb_long(&s->gb);
if(first_mb_in_slice == 0){ //FIXME better field boundary detection
if(h0->current_slice && FIELD_PICTURE){
s->chroma_y_shift = h->sps.chroma_format_idc <= 1; // 400 uses yuv420p
- s->width = 16*s->mb_width - (2>>CHROMA444)*FFMIN(h->sps.crop_right, (8<<CHROMA444)-1);
- if(h->sps.frame_mbs_only_flag)
- s->height= 16*s->mb_height - (1<<s->chroma_y_shift)*FFMIN(h->sps.crop_bottom, (16>>s->chroma_y_shift)-1);
- else
- s->height= 16*s->mb_height - (2<<s->chroma_y_shift)*FFMIN(h->sps.crop_bottom, (16>>s->chroma_y_shift)-1);
+ s->width = 16*s->mb_width;
+ s->height= 16*s->mb_height;
if (s->context_initialized
- && ( s->width != s->avctx->width || s->height != s->avctx->height
+ && ( s->width != s->avctx->coded_width || s->height != s->avctx->coded_height
+ || s->avctx->bits_per_raw_sample != h->sps.bit_depth_luma
+ || h->cur_chroma_format_idc != h->sps.chroma_format_idc
|| av_cmp_q(h->sps.sar, s->avctx->sample_aspect_ratio))) {
if(h != h0) {
- av_log_missing_feature(s->avctx, "Width/height changing with threads is", 0);
+ av_log_missing_feature(s->avctx, "Width/height/bit depth/chroma idc changing with threads is", 0);
return -1; // width / height changed during parallelized decoding
}
free_tables(h, 0);
flush_dpb(s->avctx);
MPV_common_end(s);
+ h->list_count = 0;
}
if (!s->context_initialized) {
if (h != h0) {
av_log(h->s.avctx, AV_LOG_ERROR, "Cannot (re-)initialize context during parallel decoding.\n");
return -1;
}
-
avcodec_set_dimensions(s->avctx, s->width, s->height);
+ s->avctx->width -= (2>>CHROMA444)*FFMIN(h->sps.crop_right, (8<<CHROMA444)-1);
+ s->avctx->height -= (1<<s->chroma_y_shift)*FFMIN(h->sps.crop_bottom, (16>>s->chroma_y_shift)-1) * (2 - h->sps.frame_mbs_only_flag);
s->avctx->sample_aspect_ratio= h->sps.sar;
av_assert0(s->avctx->sample_aspect_ratio.den);
+ if (s->avctx->bits_per_raw_sample != h->sps.bit_depth_luma ||
+ h->cur_chroma_format_idc != h->sps.chroma_format_idc) {
+ if (h->sps.bit_depth_luma >= 8 && h->sps.bit_depth_luma <= 10 &&
+ (h->sps.bit_depth_luma != 9 || !CHROMA422)) {
+ s->avctx->bits_per_raw_sample = h->sps.bit_depth_luma;
+ h->cur_chroma_format_idc = h->sps.chroma_format_idc;
+ h->pixel_shift = h->sps.bit_depth_luma > 8;
+
+ ff_h264dsp_init(&h->h264dsp, h->sps.bit_depth_luma, h->sps.chroma_format_idc);
+ ff_h264_pred_init(&h->hpc, s->codec_id, h->sps.bit_depth_luma, h->sps.chroma_format_idc);
+ s->dsp.dct_bits = h->sps.bit_depth_luma > 8 ? 32 : 16;
+ dsputil_init(&s->dsp, s->avctx);
+ } else {
+ av_log(s->avctx, AV_LOG_ERROR, "Unsupported bit depth: %d chroma_idc: %d\n",
+ h->sps.bit_depth_luma, h->sps.chroma_format_idc);
+ return -1;
+ }
+ }
+
if(h->sps.video_signal_type_present_flag){
- s->avctx->color_range = h->sps.full_range ? AVCOL_RANGE_JPEG : AVCOL_RANGE_MPEG;
+ s->avctx->color_range = h->sps.full_range>0 ? AVCOL_RANGE_JPEG : AVCOL_RANGE_MPEG;
if(h->sps.colour_description_present_flag){
s->avctx->color_primaries = h->sps.color_primaries;
s->avctx->color_trc = h->sps.color_trc;
switch (h->sps.bit_depth_luma) {
case 9 :
- if (CHROMA444)
- s->avctx->pix_fmt = PIX_FMT_YUV444P9;
- else if (CHROMA422)
+ if (CHROMA444) {
+ if (s->avctx->colorspace == AVCOL_SPC_RGB) {
+ s->avctx->pix_fmt = PIX_FMT_GBRP9;
+ } else
+ s->avctx->pix_fmt = PIX_FMT_YUV444P9;
+ } else if (CHROMA422)
s->avctx->pix_fmt = PIX_FMT_YUV422P9;
else
s->avctx->pix_fmt = PIX_FMT_YUV420P9;
break;
case 10 :
- if (CHROMA444)
- s->avctx->pix_fmt = PIX_FMT_YUV444P10;
- else if (CHROMA422)
+ if (CHROMA444) {
+ if (s->avctx->colorspace == AVCOL_SPC_RGB) {
+ s->avctx->pix_fmt = PIX_FMT_GBRP10;
+ } else
+ s->avctx->pix_fmt = PIX_FMT_YUV444P10;
+ } else if (CHROMA422)
s->avctx->pix_fmt = PIX_FMT_YUV422P10;
else
s->avctx->pix_fmt = PIX_FMT_YUV420P10;
break;
default:
if (CHROMA444){
+ s->avctx->pix_fmt = s->avctx->color_range == AVCOL_RANGE_JPEG ? PIX_FMT_YUVJ444P : PIX_FMT_YUV444P;
if (s->avctx->colorspace == AVCOL_SPC_RGB) {
- s->avctx->pix_fmt = PIX_FMT_GBRP;
- } else
- s->avctx->pix_fmt = s->avctx->color_range == AVCOL_RANGE_JPEG ? PIX_FMT_YUVJ444P : PIX_FMT_YUV444P;
+ s->avctx->pix_fmt = PIX_FMT_GBR24P;
+ av_log(h->s.avctx, AV_LOG_DEBUG, "Detected GBR colorspace.\n");
+ } else if (s->avctx->colorspace == AVCOL_SPC_YCGCO) {
+ av_log(h->s.avctx, AV_LOG_WARNING, "Detected unsupported YCgCo colorspace.\n");
+ }
} else if (CHROMA422) {
s->avctx->pix_fmt = s->avctx->color_range == AVCOL_RANGE_JPEG ? PIX_FMT_YUVJ422P : PIX_FMT_YUV422P;
}else{
c->sps = h->sps;
c->pps = h->pps;
c->pixel_shift = h->pixel_shift;
+ c->cur_chroma_format_idc = h->cur_chroma_format_idc;
init_scan_tables(c);
clone_tables(c, h, i);
}
if(h->sps.frame_mbs_only_flag){
s->picture_structure= PICT_FRAME;
}else{
+ if(!h->sps.direct_8x8_inference_flag && slice_type == AV_PICTURE_TYPE_B){
+ av_log(h->s.avctx, AV_LOG_ERROR, "This stream was generated by a broken encoder, invalid 8x8 inference\n");
+ return -1;
+ }
if(get_bits1(&s->gb)) { //field_pic_flag
s->picture_structure= PICT_TOP_FIELD + get_bits1(&s->gb); //bottom_field_flag
} else {
if(h0->current_slice == 0){
// Shorten frame num gaps so we don't have to allocate reference frames just to throw them away
- if(h->frame_num != h->prev_frame_num) {
+ if(h->frame_num != h->prev_frame_num && h->prev_frame_num >= 0) {
int unwrap_prev_frame_num = h->prev_frame_num, max_frame_num = 1<<h->sps.log2_max_frame_num;
if (unwrap_prev_frame_num > h->frame_num) unwrap_prev_frame_num -= max_frame_num;
}
}
- while(h->frame_num != h->prev_frame_num &&
+ while(h->frame_num != h->prev_frame_num && h->prev_frame_num >= 0 &&
h->frame_num != (h->prev_frame_num+1)%(1<<h->sps.log2_max_frame_num)){
Picture *prev = h->short_ref_count ? h->short_ref[0] : NULL;
av_log(h->s.avctx, AV_LOG_DEBUG, "Frame num gap %d %d\n", h->frame_num, h->prev_frame_num);
s0->first_field = FIELD_PICTURE;
} else {
- if (h->nal_ref_idc &&
- s0->current_picture_ptr->f.reference &&
- s0->current_picture_ptr->frame_num != h->frame_num) {
+ if (s0->current_picture_ptr->frame_num != h->frame_num) {
/*
- * This and previous field were reference, but had
+ * This and previous field had
* different frame_nums. Consider this field first in
* pair. Throw away previous field except for reference
* purposes.
h->ref_count[1]= h->pps.ref_count[1];
if(h->slice_type_nos != AV_PICTURE_TYPE_I){
+ unsigned max= (16<<(s->picture_structure != PICT_FRAME))-1;
if(h->slice_type_nos == AV_PICTURE_TYPE_B){
h->direct_spatial_mv_pred= get_bits1(&s->gb);
}
if(h->slice_type_nos==AV_PICTURE_TYPE_B)
h->ref_count[1]= get_ue_golomb(&s->gb) + 1;
- if(h->ref_count[0]-1 > 32-1 || h->ref_count[1]-1 > 32-1){
- av_log(h->s.avctx, AV_LOG_ERROR, "reference overflow\n");
- h->ref_count[0]= h->ref_count[1]= 1;
- return -1;
- }
+ }
+ if(h->ref_count[0]-1 > max || h->ref_count[1]-1 > max){
+ av_log(h->s.avctx, AV_LOG_ERROR, "reference overflow\n");
+ h->ref_count[0]= h->ref_count[1]= 1;
+ return -1;
}
if(h->slice_type_nos == AV_PICTURE_TYPE_B)
h->list_count= 2;
else
h->list_count= 1;
}else
- h->list_count= 0;
+ h->ref_count[1]= h->ref_count[0]= h->list_count= 0;
if(!default_ref_list_done){
ff_h264_fill_default_ref_list(h);
h0->last_slice_type = slice_type;
h->slice_num = ++h0->current_slice;
- if(h->slice_num >= MAX_SLICES){
- av_log(s->avctx, AV_LOG_ERROR, "Too many slices, increase MAX_SLICES and recompile\n");
+
+ if(h->slice_num)
+ h0->slice_row[(h->slice_num-1)&(MAX_SLICES-1)]= s->resync_mb_y;
+ if ( h0->slice_row[h->slice_num&(MAX_SLICES-1)] + 3 >= s->resync_mb_y
+ && h0->slice_row[h->slice_num&(MAX_SLICES-1)] <= s->resync_mb_y
+ && h->slice_num >= MAX_SLICES) {
+ //in case of ASO this check needs to be updated depending on how we decide to assign slice numbers in this case
+ av_log(s->avctx, AV_LOG_WARNING, "Possibly too many slices (%d >= %d), increase MAX_SLICES and recompile if there are artifacts\n", h->slice_num, MAX_SLICES);
}
for(j=0; j<2; j++){
if(s->mb_y >= s->mb_height){
tprintf(s->avctx, "slice end %d %d\n", get_bits_count(&s->gb), s->gb.size_in_bits);
- if(get_bits_count(&s->gb) == s->gb.size_in_bits ) {
+ if( get_bits_count(&s->gb) == s->gb.size_in_bits
+ || get_bits_count(&s->gb) < s->gb.size_in_bits && s->avctx->error_recognition < FF_ER_AGGRESSIVE) {
ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
return 0;
hx = h->thread_context[i];
hx->s.error_recognition = avctx->error_recognition;
hx->s.error_count = 0;
+ hx->x264_build= h->x264_build;
}
avctx->execute(avctx, (void *)decode_slice,
s->workaround_bugs |= FF_BUG_TRUNCATED;
if(!(s->workaround_bugs & FF_BUG_TRUNCATED)){
- while(ptr[dst_length - 1] == 0 && dst_length > 0)
+ while(dst_length > 0 && ptr[dst_length - 1] == 0)
dst_length--;
}
bit_length= !dst_length ? 0 : (8*dst_length - ff_h264_decode_rbsp_trailing(h, ptr + dst_length - 1));
if(s->avctx->debug&FF_DEBUG_STARTCODE){
- av_log(h->s.avctx, AV_LOG_DEBUG, "NAL %d at %d/%d length %d\n", hx->nal_unit_type, buf_index, buf_size, dst_length);
+ av_log(h->s.avctx, AV_LOG_DEBUG, "NAL %d/%d at %d/%d length %d\n", hx->nal_unit_type, hx->nal_ref_idc, buf_index, buf_size, dst_length);
}
if (h->is_avc && (nalsize != consumed) && nalsize){
if((err = decode_slice_header(hx, h)))
break;
+ if ( h->sei_recovery_frame_cnt >= 0
+ && ((h->recovery_frame - h->frame_num) & ((1 << h->sps.log2_max_frame_num)-1)) > h->sei_recovery_frame_cnt) {
+ h->recovery_frame = (h->frame_num + h->sei_recovery_frame_cnt) %
+ (1 << h->sps.log2_max_frame_num);
+ }
+
s->current_picture_ptr->f.key_frame |=
- (hx->nal_unit_type == NAL_IDR_SLICE) ||
- (h->sei_recovery_frame_cnt >= 0);
+ (hx->nal_unit_type == NAL_IDR_SLICE);
+
+ if (h->recovery_frame == h->frame_num) {
+ h->sync |= 1;
+ h->recovery_frame = -1;
+ }
+
+ h->sync |= !!s->current_picture_ptr->f.key_frame;
+ h->sync |= 3*!!(s->flags2 & CODEC_FLAG2_SHOW_ALL);
+ s->current_picture_ptr->sync = h->sync;
if (h->current_slice == 1) {
if(!(s->flags2 & CODEC_FLAG2_CHUNKS)) {
break;
case NAL_SPS:
init_get_bits(&s->gb, ptr, bit_length);
- ff_h264_decode_seq_parameter_set(h);
+ if(ff_h264_decode_seq_parameter_set(h) < 0 && (h->is_avc ? (nalsize != consumed) && nalsize : 1)){
+ av_log(h->s.avctx, AV_LOG_DEBUG, "SPS decoding failure, trying alternative mode\n");
+ if(h->is_avc) av_assert0(next_avc - buf_index + consumed == nalsize);
+ init_get_bits(&s->gb, &buf[buf_index + 1 - consumed], 8*(next_avc - buf_index + consumed));
+ ff_h264_decode_seq_parameter_set(h);
+ }
if (s->flags& CODEC_FLAG_LOW_DELAY ||
(h->sps.bitstream_restriction_flag && !h->sps.num_reorder_frames))
if(avctx->has_b_frames < 2)
avctx->has_b_frames= !s->low_delay;
-
- if (avctx->bits_per_raw_sample != h->sps.bit_depth_luma ||
- h->cur_chroma_format_idc != h->sps.chroma_format_idc) {
- if (h->sps.bit_depth_luma >= 8 && h->sps.bit_depth_luma <= 10) {
- avctx->bits_per_raw_sample = h->sps.bit_depth_luma;
- h->cur_chroma_format_idc = h->sps.chroma_format_idc;
- h->pixel_shift = h->sps.bit_depth_luma > 8;
-
- ff_h264dsp_init(&h->h264dsp, h->sps.bit_depth_luma, h->sps.chroma_format_idc);
- ff_h264_pred_init(&h->hpc, s->codec_id, h->sps.bit_depth_luma, h->sps.chroma_format_idc);
- s->dsp.dct_bits = h->sps.bit_depth_luma > 8 ? 32 : 16;
- dsputil_init(&s->dsp, s->avctx);
- } else {
- av_log(avctx, AV_LOG_ERROR, "Unsupported bit depth: %d\n", h->sps.bit_depth_luma);
- return -1;
- }
- }
break;
case NAL_PPS:
init_get_bits(&s->gb, ptr, bit_length);
return 0;
}
+ if(h->is_avc && buf_size >= 9 && AV_RB32(buf)==0x0164001F && buf[5] && buf[8]==0x67)
+ return ff_h264_decode_extradata(h, buf, buf_size);
buf_index=decode_nal_units(h, buf, buf_size);
if(buf_index < 0)
field_end(h, 0);
- if (!h->next_output_pic) {
- /* Wait for second field. */
- *data_size = 0;
-
- } else {
- *data_size = sizeof(AVFrame);
- *pict = *(AVFrame*)h->next_output_pic;
+ *data_size = 0; /* Wait for second field. */
+ if (h->next_output_pic && h->next_output_pic->sync) {
+ if(h->sync>1 || h->next_output_pic->f.pict_type != AV_PICTURE_TYPE_B){
+ *data_size = sizeof(AVFrame);
+ *pict = *(AVFrame*)h->next_output_pic;
+ }
}
}
#undef random
#define COUNT 8000
#define SIZE (COUNT*40)
+extern AVCodec ff_h264_decoder;
int main(void){
int i;
uint8_t temp[SIZE];
DSPContext dsp;
AVCodecContext avctx;
+ avcodec_get_context_defaults3(&avctx, &ff_h264_decoder);
+
dsputil_init(&dsp, &avctx);
init_put_bits(&pb, temp, SIZE);
{ FF_PROFILE_UNKNOWN },
};
+static const AVOption h264_options[] = {
+ {"is_avc", "is avc", offsetof(H264Context, is_avc), FF_OPT_TYPE_INT, {.dbl = 0}, 0, 1, 0},
+ {"nal_length_size", "nal_length_size", offsetof(H264Context, nal_length_size), FF_OPT_TYPE_INT, {.dbl = 0}, 0, 4, 0},
+ {NULL}
+};
+
+static const AVClass h264_class = {
+ "H264 Decoder",
+ av_default_item_name,
+ h264_options,
+ LIBAVUTIL_VERSION_INT,
+};
+
+static const AVClass h264_vdpau_class = {
+ "H264 VDPAU Decoder",
+ av_default_item_name,
+ h264_options,
+ LIBAVUTIL_VERSION_INT,
+};
+
AVCodec ff_h264_decoder = {
.name = "h264",
.type = AVMEDIA_TYPE_VIDEO,
.init_thread_copy = ONLY_IF_THREADS_ENABLED(decode_init_thread_copy),
.update_thread_context = ONLY_IF_THREADS_ENABLED(decode_update_thread_context),
.profiles = NULL_IF_CONFIG_SMALL(profiles),
+ .priv_class = &h264_class,
};
#if CONFIG_H264_VDPAU_DECODER
.long_name = NULL_IF_CONFIG_SMALL("H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10 (VDPAU acceleration)"),
.pix_fmts = (const enum PixelFormat[]){PIX_FMT_VDPAU_H264, PIX_FMT_NONE},
.profiles = NULL_IF_CONFIG_SMALL(profiles),
+ .priv_class = &h264_vdpau_class,
};
#endif
/*
* Copyright (C) 2004 Michael Niedermayer <michaelni@gmx.at>
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * Libav is distributed in the hope that it will be useful,
+ * FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "dsputil.h"
#include "dwt.h"
#include "snow.h"
+ #include "snowdata.h"
#include "rangecoder.h"
#include "mathops.h"
-
- #include "mpegvideo.h"
#include "h263.h"
#undef NDEBUG
#include <assert.h>
- static const int8_t quant3bA[256]={
- 0, 0, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
- 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
- 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
- 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
- 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
- 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
- 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
- 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
- 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
- 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
- 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
- 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
- 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
- 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
- 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
- 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
- };
-
- static const uint8_t obmc32[1024]={
- 0, 0, 0, 0, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8, 8, 4, 4, 4, 4, 4, 4, 4, 4, 0, 0, 0, 0,
- 0, 4, 4, 4, 8, 8, 8, 12, 12, 16, 16, 16, 20, 20, 20, 24, 24, 20, 20, 20, 16, 16, 16, 12, 12, 8, 8, 8, 4, 4, 4, 0,
- 0, 4, 8, 8, 12, 12, 16, 20, 20, 24, 28, 28, 32, 32, 36, 40, 40, 36, 32, 32, 28, 28, 24, 20, 20, 16, 12, 12, 8, 8, 4, 0,
- 0, 4, 8, 12, 16, 20, 24, 28, 28, 32, 36, 40, 44, 48, 52, 56, 56, 52, 48, 44, 40, 36, 32, 28, 28, 24, 20, 16, 12, 8, 4, 0,
- 4, 8, 12, 16, 20, 24, 28, 32, 40, 44, 48, 52, 56, 60, 64, 68, 68, 64, 60, 56, 52, 48, 44, 40, 32, 28, 24, 20, 16, 12, 8, 4,
- 4, 8, 12, 20, 24, 32, 36, 40, 48, 52, 56, 64, 68, 76, 80, 84, 84, 80, 76, 68, 64, 56, 52, 48, 40, 36, 32, 24, 20, 12, 8, 4,
- 4, 8, 16, 24, 28, 36, 44, 48, 56, 60, 68, 76, 80, 88, 96,100,100, 96, 88, 80, 76, 68, 60, 56, 48, 44, 36, 28, 24, 16, 8, 4,
- 4, 12, 20, 28, 32, 40, 48, 56, 64, 72, 80, 88, 92,100,108,116,116,108,100, 92, 88, 80, 72, 64, 56, 48, 40, 32, 28, 20, 12, 4,
- 4, 12, 20, 28, 40, 48, 56, 64, 72, 80, 88, 96,108,116,124,132,132,124,116,108, 96, 88, 80, 72, 64, 56, 48, 40, 28, 20, 12, 4,
- 4, 16, 24, 32, 44, 52, 60, 72, 80, 92,100,108,120,128,136,148,148,136,128,120,108,100, 92, 80, 72, 60, 52, 44, 32, 24, 16, 4,
- 4, 16, 28, 36, 48, 56, 68, 80, 88,100,112,120,132,140,152,164,164,152,140,132,120,112,100, 88, 80, 68, 56, 48, 36, 28, 16, 4,
- 4, 16, 28, 40, 52, 64, 76, 88, 96,108,120,132,144,156,168,180,180,168,156,144,132,120,108, 96, 88, 76, 64, 52, 40, 28, 16, 4,
- 8, 20, 32, 44, 56, 68, 80, 92,108,120,132,144,156,168,180,192,192,180,168,156,144,132,120,108, 92, 80, 68, 56, 44, 32, 20, 8,
- 8, 20, 32, 48, 60, 76, 88,100,116,128,140,156,168,184,196,208,208,196,184,168,156,140,128,116,100, 88, 76, 60, 48, 32, 20, 8,
- 8, 20, 36, 52, 64, 80, 96,108,124,136,152,168,180,196,212,224,224,212,196,180,168,152,136,124,108, 96, 80, 64, 52, 36, 20, 8,
- 8, 24, 40, 56, 68, 84,100,116,132,148,164,180,192,208,224,240,240,224,208,192,180,164,148,132,116,100, 84, 68, 56, 40, 24, 8,
- 8, 24, 40, 56, 68, 84,100,116,132,148,164,180,192,208,224,240,240,224,208,192,180,164,148,132,116,100, 84, 68, 56, 40, 24, 8,
- 8, 20, 36, 52, 64, 80, 96,108,124,136,152,168,180,196,212,224,224,212,196,180,168,152,136,124,108, 96, 80, 64, 52, 36, 20, 8,
- 8, 20, 32, 48, 60, 76, 88,100,116,128,140,156,168,184,196,208,208,196,184,168,156,140,128,116,100, 88, 76, 60, 48, 32, 20, 8,
- 8, 20, 32, 44, 56, 68, 80, 92,108,120,132,144,156,168,180,192,192,180,168,156,144,132,120,108, 92, 80, 68, 56, 44, 32, 20, 8,
- 4, 16, 28, 40, 52, 64, 76, 88, 96,108,120,132,144,156,168,180,180,168,156,144,132,120,108, 96, 88, 76, 64, 52, 40, 28, 16, 4,
- 4, 16, 28, 36, 48, 56, 68, 80, 88,100,112,120,132,140,152,164,164,152,140,132,120,112,100, 88, 80, 68, 56, 48, 36, 28, 16, 4,
- 4, 16, 24, 32, 44, 52, 60, 72, 80, 92,100,108,120,128,136,148,148,136,128,120,108,100, 92, 80, 72, 60, 52, 44, 32, 24, 16, 4,
- 4, 12, 20, 28, 40, 48, 56, 64, 72, 80, 88, 96,108,116,124,132,132,124,116,108, 96, 88, 80, 72, 64, 56, 48, 40, 28, 20, 12, 4,
- 4, 12, 20, 28, 32, 40, 48, 56, 64, 72, 80, 88, 92,100,108,116,116,108,100, 92, 88, 80, 72, 64, 56, 48, 40, 32, 28, 20, 12, 4,
- 4, 8, 16, 24, 28, 36, 44, 48, 56, 60, 68, 76, 80, 88, 96,100,100, 96, 88, 80, 76, 68, 60, 56, 48, 44, 36, 28, 24, 16, 8, 4,
- 4, 8, 12, 20, 24, 32, 36, 40, 48, 52, 56, 64, 68, 76, 80, 84, 84, 80, 76, 68, 64, 56, 52, 48, 40, 36, 32, 24, 20, 12, 8, 4,
- 4, 8, 12, 16, 20, 24, 28, 32, 40, 44, 48, 52, 56, 60, 64, 68, 68, 64, 60, 56, 52, 48, 44, 40, 32, 28, 24, 20, 16, 12, 8, 4,
- 0, 4, 8, 12, 16, 20, 24, 28, 28, 32, 36, 40, 44, 48, 52, 56, 56, 52, 48, 44, 40, 36, 32, 28, 28, 24, 20, 16, 12, 8, 4, 0,
- 0, 4, 8, 8, 12, 12, 16, 20, 20, 24, 28, 28, 32, 32, 36, 40, 40, 36, 32, 32, 28, 28, 24, 20, 20, 16, 12, 12, 8, 8, 4, 0,
- 0, 4, 4, 4, 8, 8, 8, 12, 12, 16, 16, 16, 20, 20, 20, 24, 24, 20, 20, 20, 16, 16, 16, 12, 12, 8, 8, 8, 4, 4, 4, 0,
- 0, 0, 0, 0, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8, 8, 4, 4, 4, 4, 4, 4, 4, 4, 0, 0, 0, 0,
- //error:0.000020
- };
- static const uint8_t obmc16[256]={
- 0, 4, 4, 8, 8, 12, 12, 16, 16, 12, 12, 8, 8, 4, 4, 0,
- 4, 8, 16, 20, 28, 32, 40, 44, 44, 40, 32, 28, 20, 16, 8, 4,
- 4, 16, 24, 36, 44, 56, 64, 76, 76, 64, 56, 44, 36, 24, 16, 4,
- 8, 20, 36, 48, 64, 76, 92,104,104, 92, 76, 64, 48, 36, 20, 8,
- 8, 28, 44, 64, 80,100,116,136,136,116,100, 80, 64, 44, 28, 8,
- 12, 32, 56, 76,100,120,144,164,164,144,120,100, 76, 56, 32, 12,
- 12, 40, 64, 92,116,144,168,196,196,168,144,116, 92, 64, 40, 12,
- 16, 44, 76,104,136,164,196,224,224,196,164,136,104, 76, 44, 16,
- 16, 44, 76,104,136,164,196,224,224,196,164,136,104, 76, 44, 16,
- 12, 40, 64, 92,116,144,168,196,196,168,144,116, 92, 64, 40, 12,
- 12, 32, 56, 76,100,120,144,164,164,144,120,100, 76, 56, 32, 12,
- 8, 28, 44, 64, 80,100,116,136,136,116,100, 80, 64, 44, 28, 8,
- 8, 20, 36, 48, 64, 76, 92,104,104, 92, 76, 64, 48, 36, 20, 8,
- 4, 16, 24, 36, 44, 56, 64, 76, 76, 64, 56, 44, 36, 24, 16, 4,
- 4, 8, 16, 20, 28, 32, 40, 44, 44, 40, 32, 28, 20, 16, 8, 4,
- 0, 4, 4, 8, 8, 12, 12, 16, 16, 12, 12, 8, 8, 4, 4, 0,
- //error:0.000015
- };
-
- //linear *64
- static const uint8_t obmc8[64]={
- 4, 12, 20, 28, 28, 20, 12, 4,
- 12, 36, 60, 84, 84, 60, 36, 12,
- 20, 60,100,140,140,100, 60, 20,
- 28, 84,140,196,196,140, 84, 28,
- 28, 84,140,196,196,140, 84, 28,
- 20, 60,100,140,140,100, 60, 20,
- 12, 36, 60, 84, 84, 60, 36, 12,
- 4, 12, 20, 28, 28, 20, 12, 4,
- //error:0.000000
- };
-
- //linear *64
- static const uint8_t obmc4[16]={
- 16, 48, 48, 16,
- 48,144,144, 48,
- 48,144,144, 48,
- 16, 48, 48, 16,
- //error:0.000000
- };
-
- static const uint8_t * const obmc_tab[4]={
- obmc32, obmc16, obmc8, obmc4
- };
-
- static int scale_mv_ref[MAX_REF_FRAMES][MAX_REF_FRAMES];
-
- typedef struct BlockNode{
- int16_t mx;
- int16_t my;
- uint8_t ref;
- uint8_t color[3];
- uint8_t type;
- //#define TYPE_SPLIT 1
- #define BLOCK_INTRA 1
- #define BLOCK_OPT 2
- //#define TYPE_NOCOLOR 4
- uint8_t level; //FIXME merge into type?
- }BlockNode;
-
- static const BlockNode null_block= { //FIXME add border maybe
- .color= {128,128,128},
- .mx= 0,
- .my= 0,
- .ref= 0,
- .type= 0,
- .level= 0,
- };
-
- #define LOG2_MB_SIZE 4
- #define MB_SIZE (1<<LOG2_MB_SIZE)
- #define ENCODER_EXTRA_BITS 4
- #define HTAPS_MAX 8
-
- typedef struct x_and_coeff{
- int16_t x;
- uint16_t coeff;
- } x_and_coeff;
-
- typedef struct SubBand{
- int level;
- int stride;
- int width;
- int height;
- int qlog; ///< log(qscale)/log[2^(1/6)]
- DWTELEM *buf;
- IDWTELEM *ibuf;
- int buf_x_offset;
- int buf_y_offset;
- int stride_line; ///< Stride measured in lines, not pixels.
- x_and_coeff * x_coeff;
- struct SubBand *parent;
- uint8_t state[/*7*2*/ 7 + 512][32];
- }SubBand;
-
- typedef struct Plane{
- int width;
- int height;
- SubBand band[MAX_DECOMPOSITIONS][4];
-
- int htaps;
- int8_t hcoeff[HTAPS_MAX/2];
- int diag_mc;
- int fast_mc;
-
- int last_htaps;
- int8_t last_hcoeff[HTAPS_MAX/2];
- int last_diag_mc;
- }Plane;
-
- typedef struct SnowContext{
- AVClass *class;
- AVCodecContext *avctx;
- RangeCoder c;
- DSPContext dsp;
- DWTContext dwt;
- AVFrame new_picture;
- AVFrame input_picture; ///< new_picture with the internal linesizes
- AVFrame current_picture;
- AVFrame last_picture[MAX_REF_FRAMES];
- uint8_t *halfpel_plane[MAX_REF_FRAMES][4][4];
- AVFrame mconly_picture;
- // uint8_t q_context[16];
- uint8_t header_state[32];
- uint8_t block_state[128 + 32*128];
- int keyframe;
- int always_reset;
- int version;
- int spatial_decomposition_type;
- int last_spatial_decomposition_type;
- int temporal_decomposition_type;
- int spatial_decomposition_count;
- int last_spatial_decomposition_count;
- int temporal_decomposition_count;
- int max_ref_frames;
- int ref_frames;
- int16_t (*ref_mvs[MAX_REF_FRAMES])[2];
- uint32_t *ref_scores[MAX_REF_FRAMES];
- DWTELEM *spatial_dwt_buffer;
- IDWTELEM *spatial_idwt_buffer;
- int colorspace_type;
- int chroma_h_shift;
- int chroma_v_shift;
- int spatial_scalability;
- int qlog;
- int last_qlog;
- int lambda;
- int lambda2;
- int pass1_rc;
- int mv_scale;
- int last_mv_scale;
- int qbias;
- int last_qbias;
- #define QBIAS_SHIFT 3
- int b_width;
- int b_height;
- int block_max_depth;
- int last_block_max_depth;
- Plane plane[MAX_PLANES];
- BlockNode *block;
- #define ME_CACHE_SIZE 1024
- int me_cache[ME_CACHE_SIZE];
- int me_cache_generation;
- slice_buffer sb;
- int memc_only;
-
- MpegEncContext m; // needed for motion estimation, should not be used for anything else, the idea is to eventually make the motion estimation independent of MpegEncContext, so this will be removed then (FIXME/XXX)
-
- uint8_t *scratchbuf;
- }SnowContext;
-
- #ifdef __sgi
- // Avoid a name clash on SGI IRIX
- #undef qexp
- #endif
- #define QEXPSHIFT (7-FRAC_BITS+8) //FIXME try to change this to 0
- static uint8_t qexp[QROOT];
-
- static inline void put_symbol(RangeCoder *c, uint8_t *state, int v, int is_signed){
- int i;
-
- if(v){
- const int a= FFABS(v);
- const int e= av_log2(a);
- const int el= FFMIN(e, 10);
- put_rac(c, state+0, 0);
-
- for(i=0; i<el; i++){
- put_rac(c, state+1+i, 1); //1..10
- }
- for(; i<e; i++){
- put_rac(c, state+1+9, 1); //1..10
- }
- put_rac(c, state+1+FFMIN(i,9), 0);
-
- for(i=e-1; i>=el; i--){
- put_rac(c, state+22+9, (a>>i)&1); //22..31
- }
- for(; i>=0; i--){
- put_rac(c, state+22+i, (a>>i)&1); //22..31
- }
-
- if(is_signed)
- put_rac(c, state+11 + el, v < 0); //11..21
- }else{
- put_rac(c, state+0, 1);
- }
- }
-
- static inline int get_symbol(RangeCoder *c, uint8_t *state, int is_signed){
- if(get_rac(c, state+0))
- return 0;
- else{
- int i, e, a;
- e= 0;
- while(get_rac(c, state+1 + FFMIN(e,9))){ //1..10
- e++;
- }
-
- a= 1;
- for(i=e-1; i>=0; i--){
- a += a + get_rac(c, state+22 + FFMIN(i,9)); //22..31
- }
-
- e= -(is_signed && get_rac(c, state+11 + FFMIN(e,10))); //11..21
- return (a^e)-e;
- }
- }
-
- static inline void put_symbol2(RangeCoder *c, uint8_t *state, int v, int log2){
- int i;
- int r= log2>=0 ? 1<<log2 : 1;
-
- assert(v>=0);
- assert(log2>=-4);
-
- while(v >= r){
- put_rac(c, state+4+log2, 1);
- v -= r;
- log2++;
- if(log2>0) r+=r;
- }
- put_rac(c, state+4+log2, 0);
-
- for(i=log2-1; i>=0; i--){
- put_rac(c, state+31-i, (v>>i)&1);
- }
- }
-
- static inline int get_symbol2(RangeCoder *c, uint8_t *state, int log2){
- int i;
- int r= log2>=0 ? 1<<log2 : 1;
- int v=0;
-
- assert(log2>=-4);
-
- while(get_rac(c, state+4+log2)){
- v+= r;
- log2++;
- if(log2>0) r+=r;
- }
-
- for(i=log2-1; i>=0; i--){
- v+= get_rac(c, state+31-i)<<i;
- }
-
- return v;
- }
-
- static inline void unpack_coeffs(SnowContext *s, SubBand *b, SubBand * parent, int orientation){
- const int w= b->width;
- const int h= b->height;
- int x,y;
-
- int run, runs;
- x_and_coeff *xc= b->x_coeff;
- x_and_coeff *prev_xc= NULL;
- x_and_coeff *prev2_xc= xc;
- x_and_coeff *parent_xc= parent ? parent->x_coeff : NULL;
- x_and_coeff *prev_parent_xc= parent_xc;
-
- runs= get_symbol2(&s->c, b->state[30], 0);
- if(runs-- > 0) run= get_symbol2(&s->c, b->state[1], 3);
- else run= INT_MAX;
-
- for(y=0; y<h; y++){
- int v=0;
- int lt=0, t=0, rt=0;
-
- if(y && prev_xc->x == 0){
- rt= prev_xc->coeff;
- }
- for(x=0; x<w; x++){
- int p=0;
- const int l= v;
-
- lt= t; t= rt;
-
- if(y){
- if(prev_xc->x <= x)
- prev_xc++;
- if(prev_xc->x == x + 1)
- rt= prev_xc->coeff;
- else
- rt=0;
- }
- if(parent_xc){
- if(x>>1 > parent_xc->x){
- parent_xc++;
- }
- if(x>>1 == parent_xc->x){
- p= parent_xc->coeff;
- }
- }
- if(/*ll|*/l|lt|t|rt|p){
- int context= av_log2(/*FFABS(ll) + */3*(l>>1) + (lt>>1) + (t&~1) + (rt>>1) + (p>>1));
- v=get_rac(&s->c, &b->state[0][context]);
- if(v){
- v= 2*(get_symbol2(&s->c, b->state[context + 2], context-4) + 1);
- v+=get_rac(&s->c, &b->state[0][16 + 1 + 3 + quant3bA[l&0xFF] + 3*quant3bA[t&0xFF]]);
+ void ff_snow_inner_add_yblock(const uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h,
+ int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8){
+ int y, x;
+ IDWTELEM * dst;
+ for(y=0; y<b_h; y++){
+ //FIXME ugly misuse of obmc_stride
+ const uint8_t *obmc1= obmc + y*obmc_stride;
+ const uint8_t *obmc2= obmc1+ (obmc_stride>>1);
+ const uint8_t *obmc3= obmc1+ obmc_stride*(obmc_stride>>1);
+ const uint8_t *obmc4= obmc3+ (obmc_stride>>1);
+ dst = slice_buffer_get_line(sb, src_y + y);
+ for(x=0; x<b_w; x++){
+ int v= obmc1[x] * block[3][x + y*src_stride]
+ +obmc2[x] * block[2][x + y*src_stride]
+ +obmc3[x] * block[1][x + y*src_stride]
+ +obmc4[x] * block[0][x + y*src_stride];
- xc->x=x;
- (xc++)->coeff= v;
- }
- }else{
- if(!run){
- if(runs-- > 0) run= get_symbol2(&s->c, b->state[1], 3);
- else run= INT_MAX;
- v= 2*(get_symbol2(&s->c, b->state[0 + 2], 0-4) + 1);
- v+=get_rac(&s->c, &b->state[0][16 + 1 + 3]);
-
- xc->x=x;
- (xc++)->coeff= v;
- }else{
- int max_run;
- run--;
- v=0;
-
- if(y) max_run= FFMIN(run, prev_xc->x - x - 2);
- else max_run= FFMIN(run, w-x-1);
- if(parent_xc)
- max_run= FFMIN(max_run, 2*parent_xc->x - x - 1);
- x+= max_run;
- run-= max_run;
- }
+ v <<= 8 - LOG2_OBMC_MAX;
+ if(FRAC_BITS != 8){
+ v >>= 8 - FRAC_BITS;
}
- }
- (xc++)->x= w+1; //end marker
- prev_xc= prev2_xc;
- prev2_xc= xc;
-
- if(parent_xc){
- if(y&1){
- while(parent_xc->x != parent->width+1)
- parent_xc++;
- parent_xc++;
- prev_parent_xc= parent_xc;
+ if(add){
+ v += dst[x + src_x];
+ v = (v + (1<<(FRAC_BITS-1))) >> FRAC_BITS;
+ if(v&(~255)) v= ~(v>>31);
+ dst8[x + y*src_stride] = v;
}else{
- parent_xc= prev_parent_xc;
+ dst[x + src_x] -= v;
}
}
}
-
- (xc++)->x= w+1; //end marker
- }
-
- 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 reset_contexts(SnowContext *s){ //FIXME better initial contexts
+ void ff_snow_reset_contexts(SnowContext *s){ //FIXME better initial contexts
int plane_index, level, orientation;
for(plane_index=0; plane_index<3; plane_index++){
memset(s->block_state, MID_STATE, sizeof(s->block_state));
}
- static int alloc_blocks(SnowContext *s){
+ int ff_snow_alloc_blocks(SnowContext *s){
int w= -((-s->avctx->width )>>LOG2_MB_SIZE);
int h= -((-s->avctx->height)>>LOG2_MB_SIZE);
return 0;
}
- static inline void set_blocks(SnowContext *s, int level, int x, int y, int l, int cb, int cr, int mx, int my, int ref, int type){
- 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;
- const int block_w= 1<<rem_depth;
- BlockNode block;
- int i,j;
-
- block.color[0]= l;
- block.color[1]= cb;
- block.color[2]= cr;
- block.mx= mx;
- block.my= my;
- block.ref= ref;
- block.type= type;
- block.level= level;
-
- for(j=0; j<block_w; j++){
- for(i=0; i<block_w; i++){
- s->block[index + i + j*w]= block;
- }
- }
- }
-
- static inline void init_ref(MotionEstContext *c, uint8_t *src[3], uint8_t *ref[3], uint8_t *ref2[3], int x, int y, int ref_index){
- const int offset[3]= {
- y*c-> stride + x,
- ((y*c->uvstride + x)>>1),
- ((y*c->uvstride + x)>>1),
- };
+ static void init_qexp(void){
int i;
- for(i=0; i<3; i++){
- c->src[0][i]= src [i];
- c->ref[0][i]= ref [i] + offset[i];
- }
- assert(!ref_index);
- }
-
- static inline void pred_mv(SnowContext *s, int *mx, int *my, int ref,
- const BlockNode *left, const BlockNode *top, const BlockNode *tr){
- if(s->ref_frames == 1){
- *mx = mid_pred(left->mx, top->mx, tr->mx);
- *my = mid_pred(left->my, top->my, tr->my);
- }else{
- const int *scale = scale_mv_ref[ref];
- *mx = mid_pred((left->mx * scale[left->ref] + 128) >>8,
- (top ->mx * scale[top ->ref] + 128) >>8,
- (tr ->mx * scale[tr ->ref] + 128) >>8);
- *my = mid_pred((left->my * scale[left->ref] + 128) >>8,
- (top ->my * scale[top ->ref] + 128) >>8,
- (tr ->my * scale[tr ->ref] + 128) >>8);
- }
- }
-
- static av_always_inline int same_block(BlockNode *a, BlockNode *b){
- if((a->type&BLOCK_INTRA) && (b->type&BLOCK_INTRA)){
- return !((a->color[0] - b->color[0]) | (a->color[1] - b->color[1]) | (a->color[2] - b->color[2]));
- }else{
- return !((a->mx - b->mx) | (a->my - b->my) | (a->ref - b->ref) | ((a->type ^ b->type)&BLOCK_INTRA));
- }
- }
-
- 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 decode_blocks(SnowContext *s){
- int x, y;
- int w= s->b_width;
- int h= s->b_height;
+ double v=128;
- for(y=0; y<h; y++){
- for(x=0; x<w; x++){
- decode_q_branch(s, 0, x, y);
- }
+ for(i=0; i<QROOT; i++){
+ qexp[i]= lrintf(v);
+ v *= pow(2, 1.0 / QROOT);
}
}
-
static void mc_block(Plane *p, uint8_t *dst, const uint8_t *src, int stride, int b_w, int b_h, int dx, int dy){
static const uint8_t weight[64]={
8,7,6,5,4,3,2,1,
}
}
- #define mca(dx,dy,b_w)\
- static void mc_block_hpel ## dx ## dy ## b_w(uint8_t *dst, const uint8_t *src, int stride, int h){\
- assert(h==b_w);\
- mc_block(NULL, dst, src-(HTAPS_MAX/2-1)-(HTAPS_MAX/2-1)*stride, stride, b_w, b_w, dx, dy);\
- }
-
- mca( 0, 0,16)
- mca( 8, 0,16)
- mca( 0, 8,16)
- mca( 8, 8,16)
- mca( 0, 0,8)
- mca( 8, 0,8)
- mca( 0, 8,8)
- mca( 8, 8,8)
-
- static void pred_block(SnowContext *s, uint8_t *dst, uint8_t *tmp, int stride, int sx, int sy, int b_w, int b_h, BlockNode *block, int plane_index, int w, int h){
+ void ff_snow_pred_block(SnowContext *s, uint8_t *dst, uint8_t *tmp, int stride, int sx, int sy, int b_w, int b_h, BlockNode *block, int plane_index, int w, int h){
if(block->type & BLOCK_INTRA){
int x, y;
const int color = block->color[plane_index];
}
}
- void ff_snow_inner_add_yblock(const uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h,
- int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8){
- int y, x;
- IDWTELEM * dst;
- for(y=0; y<b_h; y++){
- //FIXME ugly misuse of obmc_stride
- const uint8_t *obmc1= obmc + y*obmc_stride;
- const uint8_t *obmc2= obmc1+ (obmc_stride>>1);
- const uint8_t *obmc3= obmc1+ obmc_stride*(obmc_stride>>1);
- const uint8_t *obmc4= obmc3+ (obmc_stride>>1);
- dst = slice_buffer_get_line(sb, src_y + y);
- for(x=0; x<b_w; x++){
- int v= obmc1[x] * block[3][x + y*src_stride]
- +obmc2[x] * block[2][x + y*src_stride]
- +obmc3[x] * block[1][x + y*src_stride]
- +obmc4[x] * block[0][x + y*src_stride];
-
- v <<= 8 - LOG2_OBMC_MAX;
- if(FRAC_BITS != 8){
- v >>= 8 - FRAC_BITS;
- }
- if(add){
- v += dst[x + src_x];
- v = (v + (1<<(FRAC_BITS-1))) >> FRAC_BITS;
- if(v&(~255)) v= ~(v>>31);
- dst8[x + y*src_stride] = v;
- }else{
- dst[x + src_x] -= v;
- }
- }
- }
+ #define mca(dx,dy,b_w)\
+ static void mc_block_hpel ## dx ## dy ## b_w(uint8_t *dst, const uint8_t *src, int stride, int h){\
+ assert(h==b_w);\
+ mc_block(NULL, dst, src-(HTAPS_MAX/2-1)-(HTAPS_MAX/2-1)*stride, stride, b_w, b_w, dx, dy);\
}
- //FIXME name cleanup (b_w, block_w, b_width stuff)
- static av_always_inline void add_yblock(SnowContext *s, int sliced, slice_buffer *sb, IDWTELEM *dst, uint8_t *dst8, const uint8_t *obmc, int src_x, int src_y, int b_w, int b_h, int w, int h, int dst_stride, int src_stride, int obmc_stride, int b_x, int b_y, int add, int offset_dst, int plane_index){
- 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;
- BlockNode *lt= &s->block[b_x + b_y*b_stride];
- BlockNode *rt= lt+1;
- BlockNode *lb= lt+b_stride;
- BlockNode *rb= lb+1;
- uint8_t *block[4];
- int tmp_step= src_stride >= 7*MB_SIZE ? MB_SIZE : MB_SIZE*src_stride;
- uint8_t *tmp = s->scratchbuf;
- uint8_t *ptmp;
- int x,y;
-
- if(b_x<0){
- lt= rt;
- lb= rb;
- }else if(b_x + 1 >= b_width){
- rt= lt;
- rb= lb;
- }
- if(b_y<0){
- lt= lb;
- rt= rb;
- }else if(b_y + 1 >= b_height){
- lb= lt;
- rb= rt;
- }
+ mca( 0, 0,16)
+ mca( 8, 0,16)
+ mca( 0, 8,16)
+ mca( 8, 8,16)
+ mca( 0, 0,8)
+ mca( 8, 0,8)
+ mca( 0, 8,8)
+ mca( 8, 8,8)
- if(src_x<0){ //FIXME merge with prev & always round internal width up to *16
- obmc -= src_x;
- b_w += src_x;
- if(!sliced && !offset_dst)
- dst -= src_x;
- src_x=0;
- }else if(src_x + b_w > w){
- b_w = w - src_x;
- }
- if(src_y<0){
- obmc -= src_y*obmc_stride;
- b_h += src_y;
- if(!sliced && !offset_dst)
- dst -= src_y*dst_stride;
- src_y=0;
- }else if(src_y + b_h> h){
- b_h = h - src_y;
- }
+ av_cold int ff_snow_common_init(AVCodecContext *avctx){
+ SnowContext *s = avctx->priv_data;
+ int width, height;
+ int i, j;
- if(b_w<=0 || b_h<=0) return;
+ s->avctx= avctx;
+ s->max_ref_frames=1; //just make sure its not an invalid value in case of no initial keyframe
- assert(src_stride > 2*MB_SIZE + 5);
+ dsputil_init(&s->dsp, avctx);
+ ff_dwt_init(&s->dwt);
- if(!sliced && offset_dst)
- dst += src_x + src_y*dst_stride;
- dst8+= src_x + src_y*src_stride;
- // src += src_x + src_y*src_stride;
+ #define mcf(dx,dy)\
+ s->dsp.put_qpel_pixels_tab [0][dy+dx/4]=\
+ s->dsp.put_no_rnd_qpel_pixels_tab[0][dy+dx/4]=\
+ s->dsp.put_h264_qpel_pixels_tab[0][dy+dx/4];\
+ s->dsp.put_qpel_pixels_tab [1][dy+dx/4]=\
+ s->dsp.put_no_rnd_qpel_pixels_tab[1][dy+dx/4]=\
+ s->dsp.put_h264_qpel_pixels_tab[1][dy+dx/4];
- ptmp= tmp + 3*tmp_step;
- block[0]= ptmp;
- ptmp+=tmp_step;
- pred_block(s, block[0], tmp, src_stride, src_x, src_y, b_w, b_h, lt, plane_index, w, h);
+ mcf( 0, 0)
+ mcf( 4, 0)
+ mcf( 8, 0)
+ mcf(12, 0)
+ mcf( 0, 4)
+ mcf( 4, 4)
+ mcf( 8, 4)
+ mcf(12, 4)
+ mcf( 0, 8)
+ mcf( 4, 8)
+ mcf( 8, 8)
+ mcf(12, 8)
+ mcf( 0,12)
+ mcf( 4,12)
+ mcf( 8,12)
+ mcf(12,12)
- if(same_block(lt, rt)){
- block[1]= block[0];
- }else{
- block[1]= ptmp;
- ptmp+=tmp_step;
- pred_block(s, block[1], tmp, src_stride, src_x, src_y, b_w, b_h, rt, plane_index, w, h);
- }
+ #define mcfh(dx,dy)\
+ s->dsp.put_pixels_tab [0][dy/4+dx/8]=\
+ s->dsp.put_no_rnd_pixels_tab[0][dy/4+dx/8]=\
+ mc_block_hpel ## dx ## dy ## 16;\
+ s->dsp.put_pixels_tab [1][dy/4+dx/8]=\
+ s->dsp.put_no_rnd_pixels_tab[1][dy/4+dx/8]=\
+ mc_block_hpel ## dx ## dy ## 8;
- if(same_block(lt, lb)){
- block[2]= block[0];
- }else if(same_block(rt, lb)){
- block[2]= block[1];
- }else{
- block[2]= ptmp;
- ptmp+=tmp_step;
- pred_block(s, block[2], tmp, src_stride, src_x, src_y, b_w, b_h, lb, plane_index, w, h);
- }
+ mcfh(0, 0)
+ mcfh(8, 0)
+ mcfh(0, 8)
+ mcfh(8, 8)
- if(same_block(lt, rb) ){
- block[3]= block[0];
- }else if(same_block(rt, rb)){
- block[3]= block[1];
- }else if(same_block(lb, rb)){
- block[3]= block[2];
- }else{
- block[3]= ptmp;
- pred_block(s, block[3], tmp, src_stride, src_x, src_y, b_w, b_h, rb, plane_index, w, h);
- }
- if(sliced){
- s->dwt.inner_add_yblock(obmc, obmc_stride, block, b_w, b_h, src_x,src_y, src_stride, sb, add, dst8);
- }else{
- for(y=0; y<b_h; y++){
- //FIXME ugly misuse of obmc_stride
- const uint8_t *obmc1= obmc + y*obmc_stride;
- const uint8_t *obmc2= obmc1+ (obmc_stride>>1);
- const uint8_t *obmc3= obmc1+ obmc_stride*(obmc_stride>>1);
- const uint8_t *obmc4= obmc3+ (obmc_stride>>1);
- for(x=0; x<b_w; x++){
- int v= obmc1[x] * block[3][x + y*src_stride]
- +obmc2[x] * block[2][x + y*src_stride]
- +obmc3[x] * block[1][x + y*src_stride]
- +obmc4[x] * block[0][x + y*src_stride];
-
- v <<= 8 - LOG2_OBMC_MAX;
- if(FRAC_BITS != 8){
- v >>= 8 - FRAC_BITS;
- }
- if(add){
- v += dst[x + y*dst_stride];
- v = (v + (1<<(FRAC_BITS-1))) >> FRAC_BITS;
- if(v&(~255)) v= ~(v>>31);
- dst8[x + y*src_stride] = v;
- }else{
- dst[x + y*dst_stride] -= v;
- }
- }
- }
- }
- }
+ init_qexp();
- 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;
- }
- }
- }
+ // dec += FFMAX(s->chroma_h_shift, s->chroma_v_shift);
- return;
- }
+ width= s->avctx->width;
+ height= s->avctx->height;
- 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 av_always_inline void predict_slice(SnowContext *s, IDWTELEM *buf, 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];
- const 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++){
- for(x=0; x<w; x++){
- int v= buf[x + y*w] + (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++){
- for(x=0; x<w; x++){
- buf[x + y*w]-= 128<<FRAC_BITS;
- }
- }
- }
-
- return;
- }
-
- for(mb_x=0; mb_x<=mb_w; mb_x++){
- add_yblock(s, 0, NULL, buf, 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, 1, plane_index);
- }
- }
-
- static av_always_inline void predict_plane(SnowContext *s, IDWTELEM *buf, int plane_index, int add){
- const int mb_h= s->b_height << s->block_max_depth;
- int mb_y;
- for(mb_y=0; mb_y<=mb_h; mb_y++)
- predict_slice(s, buf, plane_index, add, mb_y);
- }
-
- 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){
- 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 void init_qexp(void){
- int i;
- double v=128;
-
- for(i=0; i<QROOT; i++){
- qexp[i]= lrintf(v);
- v *= pow(2, 1.0 / QROOT);
- }
- }
-
- static av_cold int common_init(AVCodecContext *avctx){
- SnowContext *s = avctx->priv_data;
- int width, height;
- int i, j;
-
- s->avctx= avctx;
- s->max_ref_frames=1; //just make sure its not an invalid value in case of no initial keyframe
-
- dsputil_init(&s->dsp, avctx);
- ff_dwt_init(&s->dwt);
-
- #define mcf(dx,dy)\
- s->dsp.put_qpel_pixels_tab [0][dy+dx/4]=\
- s->dsp.put_no_rnd_qpel_pixels_tab[0][dy+dx/4]=\
- s->dsp.put_h264_qpel_pixels_tab[0][dy+dx/4];\
- s->dsp.put_qpel_pixels_tab [1][dy+dx/4]=\
- s->dsp.put_no_rnd_qpel_pixels_tab[1][dy+dx/4]=\
- s->dsp.put_h264_qpel_pixels_tab[1][dy+dx/4];
-
- mcf( 0, 0)
- mcf( 4, 0)
- mcf( 8, 0)
- mcf(12, 0)
- mcf( 0, 4)
- mcf( 4, 4)
- mcf( 8, 4)
- mcf(12, 4)
- mcf( 0, 8)
- mcf( 4, 8)
- mcf( 8, 8)
- mcf(12, 8)
- mcf( 0,12)
- mcf( 4,12)
- mcf( 8,12)
- mcf(12,12)
-
- #define mcfh(dx,dy)\
- s->dsp.put_pixels_tab [0][dy/4+dx/8]=\
- s->dsp.put_no_rnd_pixels_tab[0][dy/4+dx/8]=\
- mc_block_hpel ## dx ## dy ## 16;\
- s->dsp.put_pixels_tab [1][dy/4+dx/8]=\
- s->dsp.put_no_rnd_pixels_tab[1][dy/4+dx/8]=\
- mc_block_hpel ## dx ## dy ## 8;
-
- mcfh(0, 0)
- mcfh(8, 0)
- mcfh(0, 8)
- mcfh(8, 8)
-
- if(!qexp[0])
- init_qexp();
-
- // dec += FFMAX(s->chroma_h_shift, s->chroma_v_shift);
-
- width= s->avctx->width;
- height= s->avctx->height;
-
- s->spatial_idwt_buffer= av_mallocz(width*height*sizeof(IDWTELEM));
- s->spatial_dwt_buffer= av_mallocz(width*height*sizeof(DWTELEM)); //FIXME this does not belong here
+ s->spatial_idwt_buffer= av_mallocz(width*height*sizeof(IDWTELEM));
+ s->spatial_dwt_buffer= av_mallocz(width*height*sizeof(DWTELEM)); //FIXME this does not belong here
for(i=0; i<MAX_REF_FRAMES; i++)
for(j=0; j<MAX_REF_FRAMES; j++)
return 0;
}
- static int common_init_after_header(AVCodecContext *avctx){
+ int ff_snow_common_init_after_header(AVCodecContext *avctx) {
SnowContext *s = avctx->priv_data;
int plane_index, level, orientation;
return 0;
}
- #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 */
-
#define USE_HALFPEL_PLANE 0
static void halfpel_interpol(SnowContext *s, uint8_t *halfpel[4][4], AVFrame *frame){
}
}
- static void release_buffer(AVCodecContext *avctx){
+ void ff_snow_release_buffer(AVCodecContext *avctx)
+ {
SnowContext *s = avctx->priv_data;
int i;
}
}
- static int frame_start(SnowContext *s){
+ int ff_snow_frame_start(SnowContext *s){
AVFrame tmp;
int w= s->avctx->width; //FIXME round up to x16 ?
int h= s->avctx->height;
EDGE_WIDTH/2, EDGE_WIDTH/2, EDGE_TOP | EDGE_BOTTOM);
}
- release_buffer(s->avctx);
+ ff_snow_release_buffer(s->avctx);
tmp= s->last_picture[s->max_ref_frames-1];
memmove(s->last_picture+1, s->last_picture, (s->max_ref_frames-1)*sizeof(AVFrame));
}
}
- s->current_picture.reference= 1;
+ s->current_picture.reference= 3;
if(s->avctx->get_buffer(s->avctx, &s->current_picture) < 0){
av_log(s->avctx, AV_LOG_ERROR, "get_buffer() failed\n");
return -1;
return 0;
}
- static av_cold void common_end(SnowContext *s){
+ av_cold void ff_snow_common_end(SnowContext *s)
+ {
int plane_index, level, orientation, i;
av_freep(&s->spatial_dwt_buffer);
s->avctx->release_buffer(s->avctx, &s->current_picture);
}
- static av_cold int decode_init(AVCodecContext *avctx)
- {
- avctx->pix_fmt= PIX_FMT_YUV420P;
-
- common_init(avctx);
-
- return 0;
- }
-
- 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;
- 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;
- }
-
- alloc_blocks(s);
-
- if(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();
-
- 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);
-
- 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"),
- };
-
- #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;
- }
-
- common_init(avctx);
- 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(¤t_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(¤t_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;
-
- 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){
- 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
-
- 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;
-
- 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);
-
- 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;
-
- 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 */
* Copyright (C) 2004 Michael Niedermayer <michaelni@gmx.at>
* Copyright (C) 2006 Robert Edele <yartrebo@earthlink.net>
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * Libav is distributed in the hope that it will be useful,
+ * FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "dsputil.h"
#include "dwt.h"
+ #include "rangecoder.h"
+ #include "mathops.h"
+ #include "mpegvideo.h"
+
#define MID_STATE 128
#define MAX_PLANES 4
#define LOG2_OBMC_MAX 8
#define OBMC_MAX (1<<(LOG2_OBMC_MAX))
+ typedef struct BlockNode{
+ int16_t mx;
+ int16_t my;
+ uint8_t ref;
+ uint8_t color[3];
+ uint8_t type;
+ //#define TYPE_SPLIT 1
+ #define BLOCK_INTRA 1
+ #define BLOCK_OPT 2
+ //#define TYPE_NOCOLOR 4
+ uint8_t level; //FIXME merge into type?
+ }BlockNode;
+
+ static const BlockNode null_block= { //FIXME add border maybe
+ .color= {128,128,128},
+ .mx= 0,
+ .my= 0,
+ .ref= 0,
+ .type= 0,
+ .level= 0,
+ };
+
+ #define LOG2_MB_SIZE 4
+ #define MB_SIZE (1<<LOG2_MB_SIZE)
+ #define ENCODER_EXTRA_BITS 4
+ #define HTAPS_MAX 8
+
+ typedef struct x_and_coeff{
+ int16_t x;
+ uint16_t coeff;
+ } x_and_coeff;
+
+ typedef struct SubBand{
+ int level;
+ int stride;
+ int width;
+ int height;
+ int qlog; ///< log(qscale)/log[2^(1/6)]
+ DWTELEM *buf;
+ IDWTELEM *ibuf;
+ int buf_x_offset;
+ int buf_y_offset;
+ int stride_line; ///< Stride measured in lines, not pixels.
+ x_and_coeff * x_coeff;
+ struct SubBand *parent;
+ uint8_t state[/*7*2*/ 7 + 512][32];
+ }SubBand;
+
+ typedef struct Plane{
+ int width;
+ int height;
+ SubBand band[MAX_DECOMPOSITIONS][4];
+
+ int htaps;
+ int8_t hcoeff[HTAPS_MAX/2];
+ int diag_mc;
+ int fast_mc;
+
+ int last_htaps;
+ int8_t last_hcoeff[HTAPS_MAX/2];
+ int last_diag_mc;
+ }Plane;
+
+ typedef struct SnowContext{
+ AVClass *class;
+ AVCodecContext *avctx;
+ RangeCoder c;
+ DSPContext dsp;
+ DWTContext dwt;
+ AVFrame new_picture;
+ AVFrame input_picture; ///< new_picture with the internal linesizes
+ AVFrame current_picture;
+ AVFrame last_picture[MAX_REF_FRAMES];
+ uint8_t *halfpel_plane[MAX_REF_FRAMES][4][4];
+ AVFrame mconly_picture;
+ // uint8_t q_context[16];
+ uint8_t header_state[32];
+ uint8_t block_state[128 + 32*128];
+ int keyframe;
+ int always_reset;
+ int version;
+ int spatial_decomposition_type;
+ int last_spatial_decomposition_type;
+ int temporal_decomposition_type;
+ int spatial_decomposition_count;
+ int last_spatial_decomposition_count;
+ int temporal_decomposition_count;
+ int max_ref_frames;
+ int ref_frames;
+ int16_t (*ref_mvs[MAX_REF_FRAMES])[2];
+ uint32_t *ref_scores[MAX_REF_FRAMES];
+ DWTELEM *spatial_dwt_buffer;
+ IDWTELEM *spatial_idwt_buffer;
+ int colorspace_type;
+ int chroma_h_shift;
+ int chroma_v_shift;
+ int spatial_scalability;
+ int qlog;
+ int last_qlog;
+ int lambda;
+ int lambda2;
+ int pass1_rc;
+ int mv_scale;
+ int last_mv_scale;
+ int qbias;
+ int last_qbias;
+ #define QBIAS_SHIFT 3
+ int b_width;
+ int b_height;
+ int block_max_depth;
+ int last_block_max_depth;
+ Plane plane[MAX_PLANES];
+ BlockNode *block;
+ #define ME_CACHE_SIZE 1024
+ int me_cache[ME_CACHE_SIZE];
+ int me_cache_generation;
+ slice_buffer sb;
+ int memc_only;
+
+ MpegEncContext m; // needed for motion estimation, should not be used for anything else, the idea is to eventually make the motion estimation independent of MpegEncContext, so this will be removed then (FIXME/XXX)
+
+ uint8_t *scratchbuf;
+ }SnowContext;
+
+ /* Tables */
+ extern const uint8_t * const obmc_tab[4];
+ #ifdef __sgi
+ // Avoid a name clash on SGI IRIX
+ #undef qexp
+ #endif
+ extern uint8_t qexp[QROOT];
+ extern int scale_mv_ref[MAX_REF_FRAMES][MAX_REF_FRAMES];
/* C bits used by mmx/sse2/altivec */
}
}
+ /* common code */
+
+ int ff_snow_common_init(AVCodecContext *avctx);
+ int ff_snow_common_init_after_header(AVCodecContext *avctx);
+ void ff_snow_common_end(SnowContext *s);
+ void ff_snow_release_buffer(AVCodecContext *avctx);
+ void ff_snow_reset_contexts(SnowContext *s);
+ int ff_snow_alloc_blocks(SnowContext *s);
+ int ff_snow_frame_start(SnowContext *s);
+ void ff_snow_pred_block(SnowContext *s, uint8_t *dst, uint8_t *tmp, int stride,
+ int sx, int sy, int b_w, int b_h, BlockNode *block,
+ int plane_index, int w, int h);
+ /* common inline functions */
+ //XXX doublecheck all of them should stay inlined
+
+ static inline void snow_set_blocks(SnowContext *s, int level, int x, int y, int l, int cb, int cr, int mx, int my, int ref, int type){
+ 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;
+ const int block_w= 1<<rem_depth;
+ BlockNode block;
+ int i,j;
+
+ block.color[0]= l;
+ block.color[1]= cb;
+ block.color[2]= cr;
+ block.mx= mx;
+ block.my= my;
+ block.ref= ref;
+ block.type= type;
+ block.level= level;
+
+ for(j=0; j<block_w; j++){
+ for(i=0; i<block_w; i++){
+ s->block[index + i + j*w]= block;
+ }
+ }
+ }
+
+ static inline void pred_mv(SnowContext *s, int *mx, int *my, int ref,
+ const BlockNode *left, const BlockNode *top, const BlockNode *tr){
+ if(s->ref_frames == 1){
+ *mx = mid_pred(left->mx, top->mx, tr->mx);
+ *my = mid_pred(left->my, top->my, tr->my);
+ }else{
+ const int *scale = scale_mv_ref[ref];
+ *mx = mid_pred((left->mx * scale[left->ref] + 128) >>8,
+ (top ->mx * scale[top ->ref] + 128) >>8,
+ (tr ->mx * scale[tr ->ref] + 128) >>8);
+ *my = mid_pred((left->my * scale[left->ref] + 128) >>8,
+ (top ->my * scale[top ->ref] + 128) >>8,
+ (tr ->my * scale[tr ->ref] + 128) >>8);
+ }
+ }
+
+ static av_always_inline int same_block(BlockNode *a, BlockNode *b){
+ if((a->type&BLOCK_INTRA) && (b->type&BLOCK_INTRA)){
+ return !((a->color[0] - b->color[0]) | (a->color[1] - b->color[1]) | (a->color[2] - b->color[2]));
+ }else{
+ return !((a->mx - b->mx) | (a->my - b->my) | (a->ref - b->ref) | ((a->type ^ b->type)&BLOCK_INTRA));
+ }
+ }
+
+ //FIXME name cleanup (b_w, block_w, b_width stuff)
+ //XXX should we really inline it?
+ static av_always_inline void add_yblock(SnowContext *s, int sliced, slice_buffer *sb, IDWTELEM *dst, uint8_t *dst8, const uint8_t *obmc, int src_x, int src_y, int b_w, int b_h, int w, int h, int dst_stride, int src_stride, int obmc_stride, int b_x, int b_y, int add, int offset_dst, int plane_index){
+ 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;
+ BlockNode *lt= &s->block[b_x + b_y*b_stride];
+ BlockNode *rt= lt+1;
+ BlockNode *lb= lt+b_stride;
+ BlockNode *rb= lb+1;
+ uint8_t *block[4];
+ int tmp_step= src_stride >= 7*MB_SIZE ? MB_SIZE : MB_SIZE*src_stride;
+ uint8_t *tmp = s->scratchbuf;
+ uint8_t *ptmp;
+ int x,y;
+
+ if(b_x<0){
+ lt= rt;
+ lb= rb;
+ }else if(b_x + 1 >= b_width){
+ rt= lt;
+ rb= lb;
+ }
+ if(b_y<0){
+ lt= lb;
+ rt= rb;
+ }else if(b_y + 1 >= b_height){
+ lb= lt;
+ rb= rt;
+ }
+
+ if(src_x<0){ //FIXME merge with prev & always round internal width up to *16
+ obmc -= src_x;
+ b_w += src_x;
+ if(!sliced && !offset_dst)
+ dst -= src_x;
+ src_x=0;
+ }else if(src_x + b_w > w){
+ b_w = w - src_x;
+ }
+ if(src_y<0){
+ obmc -= src_y*obmc_stride;
+ b_h += src_y;
+ if(!sliced && !offset_dst)
+ dst -= src_y*dst_stride;
+ src_y=0;
+ }else if(src_y + b_h> h){
+ b_h = h - src_y;
+ }
+
+ if(b_w<=0 || b_h<=0) return;
+
+ assert(src_stride > 2*MB_SIZE + 5);
+
+ if(!sliced && offset_dst)
+ dst += src_x + src_y*dst_stride;
+ dst8+= src_x + src_y*src_stride;
+ // src += src_x + src_y*src_stride;
+
+ ptmp= tmp + 3*tmp_step;
+ block[0]= ptmp;
+ ptmp+=tmp_step;
+ ff_snow_pred_block(s, block[0], tmp, src_stride, src_x, src_y, b_w, b_h, lt, plane_index, w, h);
+
+ if(same_block(lt, rt)){
+ block[1]= block[0];
+ }else{
+ block[1]= ptmp;
+ ptmp+=tmp_step;
+ ff_snow_pred_block(s, block[1], tmp, src_stride, src_x, src_y, b_w, b_h, rt, plane_index, w, h);
+ }
+
+ if(same_block(lt, lb)){
+ block[2]= block[0];
+ }else if(same_block(rt, lb)){
+ block[2]= block[1];
+ }else{
+ block[2]= ptmp;
+ ptmp+=tmp_step;
+ ff_snow_pred_block(s, block[2], tmp, src_stride, src_x, src_y, b_w, b_h, lb, plane_index, w, h);
+ }
+
+ if(same_block(lt, rb) ){
+ block[3]= block[0];
+ }else if(same_block(rt, rb)){
+ block[3]= block[1];
+ }else if(same_block(lb, rb)){
+ block[3]= block[2];
+ }else{
+ block[3]= ptmp;
+ ff_snow_pred_block(s, block[3], tmp, src_stride, src_x, src_y, b_w, b_h, rb, plane_index, w, h);
+ }
+ if(sliced){
+ s->dwt.inner_add_yblock(obmc, obmc_stride, block, b_w, b_h, src_x,src_y, src_stride, sb, add, dst8);
+ }else{
+ for(y=0; y<b_h; y++){
+ //FIXME ugly misuse of obmc_stride
+ const uint8_t *obmc1= obmc + y*obmc_stride;
+ const uint8_t *obmc2= obmc1+ (obmc_stride>>1);
+ const uint8_t *obmc3= obmc1+ obmc_stride*(obmc_stride>>1);
+ const uint8_t *obmc4= obmc3+ (obmc_stride>>1);
+ for(x=0; x<b_w; x++){
+ int v= obmc1[x] * block[3][x + y*src_stride]
+ +obmc2[x] * block[2][x + y*src_stride]
+ +obmc3[x] * block[1][x + y*src_stride]
+ +obmc4[x] * block[0][x + y*src_stride];
+
+ v <<= 8 - LOG2_OBMC_MAX;
+ if(FRAC_BITS != 8){
+ v >>= 8 - FRAC_BITS;
+ }
+ if(add){
+ v += dst[x + y*dst_stride];
+ v = (v + (1<<(FRAC_BITS-1))) >> FRAC_BITS;
+ if(v&(~255)) v= ~(v>>31);
+ dst8[x + y*src_stride] = v;
+ }else{
+ dst[x + y*dst_stride] -= v;
+ }
+ }
+ }
+ }
+ }
+
+ static av_always_inline void predict_slice(SnowContext *s, IDWTELEM *buf, 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];
+ const 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++){
+ for(x=0; x<w; x++){
+ int v= buf[x + y*w] + (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++){
+ for(x=0; x<w; x++){
+ buf[x + y*w]-= 128<<FRAC_BITS;
+ }
+ }
+ }
+
+ return;
+ }
+
+ for(mb_x=0; mb_x<=mb_w; mb_x++){
+ add_yblock(s, 0, NULL, buf, 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, 1, plane_index);
+ }
+ }
+
+ static av_always_inline void predict_plane(SnowContext *s, IDWTELEM *buf, int plane_index, int add){
+ const int mb_h= s->b_height << s->block_max_depth;
+ int mb_y;
+ for(mb_y=0; mb_y<=mb_h; mb_y++)
+ predict_slice(s, buf, plane_index, add, mb_y);
+ }
+
+ static inline void set_blocks(SnowContext *s, int level, int x, int y, int l, int cb, int cr, int mx, int my, int ref, int type){
+ 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;
+ const int block_w= 1<<rem_depth;
+ BlockNode block;
+ int i,j;
+
+ block.color[0]= l;
+ block.color[1]= cb;
+ block.color[2]= cr;
+ block.mx= mx;
+ block.my= my;
+ block.ref= ref;
+ block.type= type;
+ block.level= level;
+
+ for(j=0; j<block_w; j++){
+ for(i=0; i<block_w; i++){
+ s->block[index + i + j*w]= block;
+ }
+ }
+ }
+
+ static inline void init_ref(MotionEstContext *c, uint8_t *src[3], uint8_t *ref[3], uint8_t *ref2[3], int x, int y, int ref_index){
+ const int offset[3]= {
+ y*c-> stride + x,
+ ((y*c->uvstride + x)>>1),
+ ((y*c->uvstride + x)>>1),
+ };
+ int i;
+ for(i=0; i<3; i++){
+ c->src[0][i]= src [i];
+ c->ref[0][i]= ref [i] + offset[i];
+ }
+ assert(!ref_index);
+ }
+
+
+ /* bitstream functions */
+
+ extern const int8_t quant3bA[256];
+
+ #define QEXPSHIFT (7-FRAC_BITS+8) //FIXME try to change this to 0
+
+ static inline void put_symbol(RangeCoder *c, uint8_t *state, int v, int is_signed){
+ int i;
+
+ if(v){
+ const int a= FFABS(v);
+ const int e= av_log2(a);
+ const int el= FFMIN(e, 10);
+ put_rac(c, state+0, 0);
+
+ for(i=0; i<el; i++){
+ put_rac(c, state+1+i, 1); //1..10
+ }
+ for(; i<e; i++){
+ put_rac(c, state+1+9, 1); //1..10
+ }
+ put_rac(c, state+1+FFMIN(i,9), 0);
+
+ for(i=e-1; i>=el; i--){
+ put_rac(c, state+22+9, (a>>i)&1); //22..31
+ }
+ for(; i>=0; i--){
+ put_rac(c, state+22+i, (a>>i)&1); //22..31
+ }
+
+ if(is_signed)
+ put_rac(c, state+11 + el, v < 0); //11..21
+ }else{
+ put_rac(c, state+0, 1);
+ }
+ }
+
+ static inline int get_symbol(RangeCoder *c, uint8_t *state, int is_signed){
+ if(get_rac(c, state+0))
+ return 0;
+ else{
+ int i, e, a;
+ e= 0;
+ while(get_rac(c, state+1 + FFMIN(e,9))){ //1..10
+ e++;
+ }
+
+ a= 1;
+ for(i=e-1; i>=0; i--){
+ a += a + get_rac(c, state+22 + FFMIN(i,9)); //22..31
+ }
+
+ e= -(is_signed && get_rac(c, state+11 + FFMIN(e,10))); //11..21
+ return (a^e)-e;
+ }
+ }
+
+ static inline void put_symbol2(RangeCoder *c, uint8_t *state, int v, int log2){
+ int i;
+ int r= log2>=0 ? 1<<log2 : 1;
+
+ assert(v>=0);
+ assert(log2>=-4);
+
+ while(v >= r){
+ put_rac(c, state+4+log2, 1);
+ v -= r;
+ log2++;
+ if(log2>0) r+=r;
+ }
+ put_rac(c, state+4+log2, 0);
+
+ for(i=log2-1; i>=0; i--){
+ put_rac(c, state+31-i, (v>>i)&1);
+ }
+ }
+
+ static inline int get_symbol2(RangeCoder *c, uint8_t *state, int log2){
+ int i;
+ int r= log2>=0 ? 1<<log2 : 1;
+ int v=0;
+
+ assert(log2>=-4);
+
+ while(get_rac(c, state+4+log2)){
+ v+= r;
+ log2++;
+ if(log2>0) r+=r;
+ }
+
+ for(i=log2-1; i>=0; i--){
+ v+= get_rac(c, state+31-i)<<i;
+ }
+
+ return v;
+ }
+
+ static inline void unpack_coeffs(SnowContext *s, SubBand *b, SubBand * parent, int orientation){
+ const int w= b->width;
+ const int h= b->height;
+ int x,y;
+
+ int run, runs;
+ x_and_coeff *xc= b->x_coeff;
+ x_and_coeff *prev_xc= NULL;
+ x_and_coeff *prev2_xc= xc;
+ x_and_coeff *parent_xc= parent ? parent->x_coeff : NULL;
+ x_and_coeff *prev_parent_xc= parent_xc;
+
+ runs= get_symbol2(&s->c, b->state[30], 0);
+ if(runs-- > 0) run= get_symbol2(&s->c, b->state[1], 3);
+ else run= INT_MAX;
+
+ for(y=0; y<h; y++){
+ int v=0;
+ int lt=0, t=0, rt=0;
+
+ if(y && prev_xc->x == 0){
+ rt= prev_xc->coeff;
+ }
+ for(x=0; x<w; x++){
+ int p=0;
+ const int l= v;
+
+ lt= t; t= rt;
+
+ if(y){
+ if(prev_xc->x <= x)
+ prev_xc++;
+ if(prev_xc->x == x + 1)
+ rt= prev_xc->coeff;
+ else
+ rt=0;
+ }
+ if(parent_xc){
+ if(x>>1 > parent_xc->x){
+ parent_xc++;
+ }
+ if(x>>1 == parent_xc->x){
+ p= parent_xc->coeff;
+ }
+ }
+ if(/*ll|*/l|lt|t|rt|p){
+ int context= av_log2(/*FFABS(ll) + */3*(l>>1) + (lt>>1) + (t&~1) + (rt>>1) + (p>>1));
+
+ v=get_rac(&s->c, &b->state[0][context]);
+ if(v){
+ v= 2*(get_symbol2(&s->c, b->state[context + 2], context-4) + 1);
+ v+=get_rac(&s->c, &b->state[0][16 + 1 + 3 + quant3bA[l&0xFF] + 3*quant3bA[t&0xFF]]);
+
+ xc->x=x;
+ (xc++)->coeff= v;
+ }
+ }else{
+ if(!run){
+ if(runs-- > 0) run= get_symbol2(&s->c, b->state[1], 3);
+ else run= INT_MAX;
+ v= 2*(get_symbol2(&s->c, b->state[0 + 2], 0-4) + 1);
+ v+=get_rac(&s->c, &b->state[0][16 + 1 + 3]);
+
+ xc->x=x;
+ (xc++)->coeff= v;
+ }else{
+ int max_run;
+ run--;
+ v=0;
+
+ if(y) max_run= FFMIN(run, prev_xc->x - x - 2);
+ else max_run= FFMIN(run, w-x-1);
+ if(parent_xc)
+ max_run= FFMIN(max_run, 2*parent_xc->x - x - 1);
+ x+= max_run;
+ run-= max_run;
+ }
+ }
+ }
+ (xc++)->x= w+1; //end marker
+ prev_xc= prev2_xc;
+ prev2_xc= xc;
+
+ if(parent_xc){
+ if(y&1){
+ while(parent_xc->x != parent->width+1)
+ parent_xc++;
+ parent_xc++;
+ prev_parent_xc= parent_xc;
+ }else{
+ parent_xc= prev_parent_xc;
+ }
+ }
+ }
+
+ (xc++)->x= w+1; //end marker
+ }
+
#endif /* AVCODEC_SNOW_H */
--- /dev/null
- * 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"),
+ };
--- /dev/null
- * 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(¤t_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(¤t_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 */
* TIFF image encoder
* Copyright (c) 2007 Bartlomiej Wolowiec
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * Libav is distributed in the hope that it will be useful,
+ * FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#if CONFIG_ZLIB
#include <zlib.h>
#endif
+#include "libavutil/opt.h"
#include "bytestream.h"
#include "tiff.h"
#include "rle.h"
int buf_size; ///< buffer size
uint16_t subsampling[2]; ///< YUV subsampling factors
struct LZWEncodeState *lzws; ///< LZW Encode state
+ uint32_t dpi; ///< image resolution in DPI
} TiffEncoderContext;
uint32_t *strip_sizes = NULL;
uint32_t *strip_offsets = NULL;
int bytes_per_row;
- uint32_t res[2] = { 72, 1 }; // image resolution (72/1)
- static const uint16_t bpp_tab[] = { 8, 8, 8, 8 };
+ uint32_t res[2] = { s->dpi, 1 }; // image resolution (72/1)
+ uint16_t bpp_tab[] = { 8, 8, 8, 8 };
int ret = -1;
int is_yuv = 0;
uint8_t *yuv_line = NULL;
p->key_frame = 1;
avctx->coded_frame= &s->picture;
+ #if FF_API_TIFFENC_COMPLEVEL
+ if (avctx->compression_level != FF_COMPRESSION_DEFAULT)
+ av_log(avctx, AV_LOG_WARNING, "Using compression_level to set compression "
+ "algorithm is deprecated. Please use the compression_algo private "
+ "option instead.\n");
if (avctx->compression_level == 0) {
s->compr = TIFF_RAW;
} else if(avctx->compression_level == 2) {
s->compr = TIFF_DEFLATE;
#endif
}
+ #endif
s->width = avctx->width;
s->height = avctx->height;
s->subsampling[1] = 1;
switch (avctx->pix_fmt) {
+ case PIX_FMT_RGB48LE:
+ s->bpp = 48;
+ s->photometric_interpretation = 2;
+ bpp_tab[0] = 16;
+ bpp_tab[1] = 16;
+ bpp_tab[2] = 16;
+ bpp_tab[3] = 16;
+ break;
case PIX_FMT_RGB24:
s->bpp = 24;
s->photometric_interpretation = 2;
s->photometric_interpretation = 3;
break;
case PIX_FMT_MONOBLACK:
- s->bpp = 1;
- s->photometric_interpretation = 1;
- break;
case PIX_FMT_MONOWHITE:
s->bpp = 1;
- s->photometric_interpretation = 0;
+ s->photometric_interpretation = avctx->pix_fmt == PIX_FMT_MONOBLACK;
+ bpp_tab[0] = 1;
break;
case PIX_FMT_YUV420P:
case PIX_FMT_YUV422P:
return -1;
}
if (!is_yuv)
- s->bpp_tab_size = (s->bpp >> 3);
+ s->bpp_tab_size = (s->bpp >= 48) ? ((s->bpp + 7) >> 4):((s->bpp + 7) >> 3);
if (s->compr == TIFF_DEFLATE || s->compr == TIFF_ADOBE_DEFLATE || s->compr == TIFF_LZW)
//best choose for DEFLATE
#define OFFSET(x) offsetof(TiffEncoderContext, x)
#define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
static const AVOption options[] = {
+ {"dpi", "set the image resolution (in dpi)", OFFSET(dpi), AV_OPT_TYPE_INT, {.dbl = 72}, 1, 0x10000, AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_ENCODING_PARAM},
{ "compression_algo", NULL, OFFSET(compr), AV_OPT_TYPE_INT, {TIFF_PACKBITS}, TIFF_RAW, TIFF_DEFLATE, VE, "compression_algo" },
{ "packbits", NULL, 0, AV_OPT_TYPE_CONST, {TIFF_PACKBITS}, 0, 0, VE, "compression_algo" },
{ "raw", NULL, 0, AV_OPT_TYPE_CONST, {TIFF_RAW}, 0, 0, VE, "compression_algo" },
PIX_FMT_MONOBLACK, PIX_FMT_MONOWHITE,
PIX_FMT_YUV420P, PIX_FMT_YUV422P,
PIX_FMT_YUV444P, PIX_FMT_YUV410P,
- PIX_FMT_YUV411P,
+ PIX_FMT_YUV411P, PIX_FMT_RGB48LE,
PIX_FMT_NONE},
.long_name = NULL_IF_CONFIG_SMALL("TIFF image"),
.priv_class = &tiffenc_class,
* Copyright (c) 2001 Fabrice Bellard
* Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * Libav is distributed in the hope that it will be useful,
+ * FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
case PIX_FMT_YUV422P:
case PIX_FMT_YUV440P:
case PIX_FMT_YUV444P:
+ case PIX_FMT_GBRP:
case PIX_FMT_GRAY8:
case PIX_FMT_GRAY16BE:
case PIX_FMT_GRAY16LE:
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:
#if HAVE_MMX
if(s->codec_id == CODEC_ID_SVQ1 || s->codec_id == CODEC_ID_VP5 ||
s->codec_id == CODEC_ID_VP6 || s->codec_id == CODEC_ID_VP6F ||
- s->codec_id == CODEC_ID_VP6A) {
+ s->codec_id == CODEC_ID_VP6A || s->codec_id == CODEC_ID_DIRAC) {
linesize_align[0] =
linesize_align[1] =
linesize_align[2] = 16;
*width=FFALIGN(*width, align);
}
+void ff_init_buffer_info(AVCodecContext *s, AVFrame *pic)
+{
+ if (s->pkt) {
+ pic->pkt_pts = s->pkt->pts;
+ pic->pkt_pos = s->pkt->pos;
+ } else {
+ pic->pkt_pts = AV_NOPTS_VALUE;
+ pic->pkt_pos = -1;
+ }
+ pic->reordered_opaque= s->reordered_opaque;
+ pic->sample_aspect_ratio = s->sample_aspect_ratio;
+ pic->width = s->width;
+ pic->height = s->height;
+ pic->format = s->pix_fmt;
+}
+
int avcodec_default_get_buffer(AVCodecContext *s, AVFrame *pic){
int i;
int w= s->width;
}
avci->buffer_count++;
- if(s->pkt) pic->pkt_pts= s->pkt->pts;
- else pic->pkt_pts= AV_NOPTS_VALUE;
+ if (s->pkt) {
+ pic->pkt_pts = s->pkt->pts;
+ pic->pkt_pos = s->pkt->pos;
+ } else {
+ pic->pkt_pts = AV_NOPTS_VALUE;
+ pic->pkt_pos = -1;
+ }
pic->reordered_opaque= s->reordered_opaque;
+ pic->sample_aspect_ratio = s->sample_aspect_ratio;
+ pic->width = s->width;
+ pic->height = s->height;
+ pic->format = s->pix_fmt;
if(s->debug&FF_DEBUG_BUFFERS)
av_log(s, AV_LOG_DEBUG, "default_get_buffer called on pic %p, %d "
void avcodec_get_frame_defaults(AVFrame *pic){
memset(pic, 0, sizeof(AVFrame));
- pic->pts= AV_NOPTS_VALUE;
+ pic->pts = pic->best_effort_timestamp = AV_NOPTS_VALUE;
+ pic->pkt_pos = -1;
pic->key_frame= 1;
+ pic->sample_aspect_ratio = (AVRational){0, 1};
+ pic->format = -1; /* unknown */
}
AVFrame *avcodec_alloc_frame(void){
return pic;
}
+static void avcodec_get_subtitle_defaults(AVSubtitle *sub)
+{
+ memset(sub, 0, sizeof(*sub));
+ sub->pts = AV_NOPTS_VALUE;
+}
+
#if FF_API_AVCODEC_OPEN
int attribute_align_arg avcodec_open(AVCodecContext *avctx, AVCodec *codec)
{
av_log(avctx, AV_LOG_DEBUG, "err{or,}_recognition separate: %d; %d\n",
avctx->error_recognition, avctx->err_recognition);
- /* FF_ER_CAREFUL (==1) implies AV_EF_CRCCHECK (== 1<<1 - 1),
- FF_ER_COMPLIANT (==2) implies AV_EF_{CRCCHECK,BITSTREAM} (== 1<<2 - 1), et cetera} */
- avctx->err_recognition |= (1<<(avctx->error_recognition-(avctx->error_recognition>=FF_ER_VERY_AGGRESSIVE))) - 1;
+ switch(avctx->error_recognition){
+ case FF_ER_EXPLODE : avctx->err_recognition |= AV_EF_EXPLODE | AV_EF_COMPLIANT | AV_EF_CAREFUL;
+ break;
+ case FF_ER_VERY_AGGRESSIVE:
+ case FF_ER_AGGRESSIVE : avctx->err_recognition |= AV_EF_AGGRESSIVE;
+ case FF_ER_COMPLIANT : avctx->err_recognition |= AV_EF_COMPLIANT;
+ case FF_ER_CAREFUL : avctx->err_recognition |= AV_EF_CAREFUL;
+ }
+
av_log(avctx, AV_LOG_DEBUG, "err{or,}_recognition combined: %d; %d\n",
avctx->error_recognition, avctx->err_recognition);
#endif
+ if (!HAVE_THREADS)
+ av_log(avctx, AV_LOG_WARNING, "Warning: not compiled with thread support, using thread emulation\n");
+
if (HAVE_THREADS && !avctx->thread_opaque) {
ret = ff_thread_init(avctx);
if (ret < 0) {
}
}
- if (avctx->codec->max_lowres < avctx->lowres) {
+ if (avctx->codec->max_lowres < avctx->lowres || avctx->lowres < 0) {
av_log(avctx, AV_LOG_ERROR, "The maximum value for lowres supported by the decoder is %d\n",
avctx->codec->max_lowres);
ret = AVERROR(EINVAL);
}
}
+ avctx->pts_correction_num_faulty_pts =
+ avctx->pts_correction_num_faulty_dts = 0;
+ avctx->pts_correction_last_pts =
+ avctx->pts_correction_last_dts = INT64_MIN;
+
if(avctx->codec->init && !(avctx->active_thread_type&FF_THREAD_FRAME)){
ret = avctx->codec->init(avctx);
if (ret < 0) {
goto free_and_end;
}
}
+
+ ret=0;
end:
entangled_thread_counter--;
av_log(avctx, AV_LOG_ERROR, "start_display_time must be 0.\n");
return -1;
}
- if(sub->num_rects == 0 || !sub->rects)
- return -1;
+
ret = avctx->codec->encode(avctx, buf, buf_size, sub);
avctx->frame_number++;
return ret;
}
+/**
+ * Attempt to guess proper monotonic timestamps for decoded video frames
+ * which might have incorrect times. Input timestamps may wrap around, in
+ * which case the output will as well.
+ *
+ * @param pts the pts field of the decoded AVPacket, as passed through
+ * AVFrame.pkt_pts
+ * @param dts the dts field of the decoded AVPacket
+ * @return one of the input values, may be AV_NOPTS_VALUE
+ */
+static int64_t guess_correct_pts(AVCodecContext *ctx,
+ int64_t reordered_pts, int64_t dts)
+{
+ int64_t pts = AV_NOPTS_VALUE;
+
+ if (dts != AV_NOPTS_VALUE) {
+ ctx->pts_correction_num_faulty_dts += dts <= ctx->pts_correction_last_dts;
+ ctx->pts_correction_last_dts = dts;
+ }
+ if (reordered_pts != AV_NOPTS_VALUE) {
+ ctx->pts_correction_num_faulty_pts += reordered_pts <= ctx->pts_correction_last_pts;
+ ctx->pts_correction_last_pts = reordered_pts;
+ }
+ if ((ctx->pts_correction_num_faulty_pts<=ctx->pts_correction_num_faulty_dts || dts == AV_NOPTS_VALUE)
+ && reordered_pts != AV_NOPTS_VALUE)
+ pts = reordered_pts;
+ else
+ pts = dts;
+
+ return pts;
+}
+
int attribute_align_arg avcodec_decode_video2(AVCodecContext *avctx, AVFrame *picture,
int *got_picture_ptr,
AVPacket *avpkt)
if((avctx->coded_width||avctx->coded_height) && av_image_check_size(avctx->coded_width, avctx->coded_height, 0, avctx))
return -1;
- avctx->pkt = avpkt;
-
if((avctx->codec->capabilities & CODEC_CAP_DELAY) || avpkt->size || (avctx->active_thread_type&FF_THREAD_FRAME)){
+ av_packet_split_side_data(avpkt);
+ avctx->pkt = avpkt;
if (HAVE_THREADS && avctx->active_thread_type&FF_THREAD_FRAME)
ret = ff_thread_decode_frame(avctx, picture, got_picture_ptr,
avpkt);
ret = avctx->codec->decode(avctx, picture, got_picture_ptr,
avpkt);
picture->pkt_dts= avpkt->dts;
+
+ if(!avctx->has_b_frames){
+ picture->pkt_pos= avpkt->pos;
+ }
+ //FIXME these should be under if(!avctx->has_b_frames)
+ if (!picture->sample_aspect_ratio.num)
+ picture->sample_aspect_ratio = avctx->sample_aspect_ratio;
+ if (!picture->width)
+ picture->width = avctx->width;
+ if (!picture->height)
+ picture->height = avctx->height;
+ if (picture->format == PIX_FMT_NONE)
+ picture->format = avctx->pix_fmt;
}
emms_c(); //needed to avoid an emms_c() call before every return;
- if (*got_picture_ptr)
+
+ if (*got_picture_ptr){
avctx->frame_number++;
+ picture->best_effort_timestamp = guess_correct_pts(avctx,
+ picture->pkt_pts,
+ picture->pkt_dts);
+ }
}else
ret= 0;
avctx->pkt = avpkt;
*got_sub_ptr = 0;
+ avcodec_get_subtitle_defaults(sub);
ret = avctx->codec->decode(avctx, sub, got_sub_ptr, avpkt);
if (*got_sub_ptr)
avctx->frame_number++;
return 0;
}
+static enum CodecID remap_deprecated_codec_id(enum CodecID id)
+{
+ switch(id){
+ case CODEC_ID_G723_1_DEPRECATED : return CODEC_ID_G723_1;
+ case CODEC_ID_G729_DEPRECATED : return CODEC_ID_G729;
+ case CODEC_ID_UTVIDEO_DEPRECATED: return CODEC_ID_UTVIDEO;
+ default : return id;
+ }
+}
+
AVCodec *avcodec_find_encoder(enum CodecID id)
{
AVCodec *p, *experimental=NULL;
p = first_avcodec;
+ id= remap_deprecated_codec_id(id);
while (p) {
if (p->encode != NULL && p->id == id) {
if (p->capabilities & CODEC_CAP_EXPERIMENTAL && !experimental) {
AVCodec *avcodec_find_decoder(enum CodecID id)
{
- AVCodec *p;
+ AVCodec *p, *experimental=NULL;
p = first_avcodec;
+ id= remap_deprecated_codec_id(id);
while (p) {
- if (p->decode != NULL && p->id == id)
- return p;
+ if (p->decode != NULL && p->id == id) {
+ if (p->capabilities & CODEC_CAP_EXPERIMENTAL && !experimental) {
+ experimental = p;
+ } else
+ return p;
+ }
p = p->next;
}
- return NULL;
+ return experimental;
}
AVCodec *avcodec_find_decoder_by_name(const char *name)
return bit_rate;
}
+const char *avcodec_get_name(enum CodecID id)
+{
+ AVCodec *codec;
+
+#if !CONFIG_SMALL
+ switch (id) {
+#include "libavcodec/codec_names.h"
+ }
+ av_log(NULL, AV_LOG_WARNING, "Codec 0x%x is not in the full list.\n", id);
+#endif
+ codec = avcodec_find_decoder(id);
+ if (codec)
+ return codec->name;
+ codec = avcodec_find_encoder(id);
+ if (codec)
+ return codec->name;
+ return "unknown_codec";
+}
+
size_t av_get_codec_tag_string(char *buf, size_t buf_size, unsigned int codec_tag)
{
int i, len, ret = 0;
void avcodec_string(char *buf, int buf_size, AVCodecContext *enc, int encode)
{
+ const char *codec_type;
const char *codec_name;
const char *profile = NULL;
AVCodec *p;
- char buf1[32];
int bitrate;
AVRational display_aspect_ratio;
- if (encode)
- p = avcodec_find_encoder(enc->codec_id);
- else
- p = avcodec_find_decoder(enc->codec_id);
-
- if (p) {
- codec_name = p->name;
- profile = av_get_profile_name(p, enc->profile);
- } else if (enc->codec_id == CODEC_ID_MPEG2TS) {
- /* fake mpeg2 transport stream codec (currently not
- registered) */
- codec_name = "mpeg2ts";
- } else if (enc->codec_name[0] != '\0') {
- codec_name = enc->codec_name;
- } else {
- /* output avi tags */
+ if (!buf || buf_size <= 0)
+ return;
+ codec_type = av_get_media_type_string(enc->codec_type);
+ codec_name = avcodec_get_name(enc->codec_id);
+ if (enc->profile != FF_PROFILE_UNKNOWN) {
+ p = encode ? avcodec_find_encoder(enc->codec_id) :
+ avcodec_find_decoder(enc->codec_id);
+ if (p)
+ profile = av_get_profile_name(p, enc->profile);
+ }
+
+ snprintf(buf, buf_size, "%s: %s%s", codec_type ? codec_type : "unknown",
+ codec_name, enc->mb_decision ? " (hq)" : "");
+ buf[0] ^= 'a' ^ 'A'; /* first letter in uppercase */
+ if (profile)
+ snprintf(buf + strlen(buf), buf_size - strlen(buf), " (%s)", profile);
+ if (enc->codec_tag) {
char tag_buf[32];
av_get_codec_tag_string(tag_buf, sizeof(tag_buf), enc->codec_tag);
- snprintf(buf1, sizeof(buf1), "%s / 0x%04X", tag_buf, enc->codec_tag);
- codec_name = buf1;
+ snprintf(buf + strlen(buf), buf_size - strlen(buf),
+ " (%s / 0x%04X)", tag_buf, enc->codec_tag);
}
-
switch(enc->codec_type) {
case AVMEDIA_TYPE_VIDEO:
- snprintf(buf, buf_size,
- "Video: %s%s",
- codec_name, enc->mb_decision ? " (hq)" : "");
- if (profile)
- snprintf(buf + strlen(buf), buf_size - strlen(buf),
- " (%s)", profile);
if (enc->pix_fmt != PIX_FMT_NONE) {
snprintf(buf + strlen(buf), buf_size - strlen(buf),
", %s",
enc->height*enc->sample_aspect_ratio.den,
1024*1024);
snprintf(buf + strlen(buf), buf_size - strlen(buf),
- " [PAR %d:%d DAR %d:%d]",
+ " [SAR %d:%d DAR %d:%d]",
enc->sample_aspect_ratio.num, enc->sample_aspect_ratio.den,
display_aspect_ratio.num, display_aspect_ratio.den);
}
}
break;
case AVMEDIA_TYPE_AUDIO:
- snprintf(buf, buf_size,
- "Audio: %s",
- codec_name);
- if (profile)
- snprintf(buf + strlen(buf), buf_size - strlen(buf),
- " (%s)", profile);
if (enc->sample_rate) {
snprintf(buf + strlen(buf), buf_size - strlen(buf),
", %d Hz", enc->sample_rate);
", %s", av_get_sample_fmt_name(enc->sample_fmt));
}
break;
- case AVMEDIA_TYPE_DATA:
- snprintf(buf, buf_size, "Data: %s", codec_name);
- break;
- case AVMEDIA_TYPE_SUBTITLE:
- snprintf(buf, buf_size, "Subtitle: %s", codec_name);
- break;
- case AVMEDIA_TYPE_ATTACHMENT:
- snprintf(buf, buf_size, "Attachment: %s", codec_name);
- break;
default:
- snprintf(buf, buf_size, "Invalid Codec type %d", enc->codec_type);
return;
}
if (encode) {
const char *avcodec_configuration(void)
{
- return LIBAV_CONFIGURATION;
+ return FFMPEG_CONFIGURATION;
}
const char *avcodec_license(void)
{
#define LICENSE_PREFIX "libavcodec license: "
- return LICENSE_PREFIX LIBAV_LICENSE + sizeof(LICENSE_PREFIX) - 1;
+ return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
}
void avcodec_flush_buffers(AVCodecContext *avctx)
void av_log_missing_feature(void *avc, const char *feature, int want_sample)
{
- av_log(avc, AV_LOG_WARNING, "%s not implemented. Update your Libav "
+ av_log(avc, AV_LOG_WARNING, "%s not implemented. Update your FFmpeg "
"version to the newest one from Git. If the problem still "
"occurs, it means that your file has a feature which has not "
"been implemented.\n", feature);
if (msg)
av_vlog(avc, AV_LOG_WARNING, msg, argument_list);
av_log(avc, AV_LOG_WARNING, "If you want to help, upload a sample "
- "of this file to ftp://upload.libav.org/incoming/ "
- "and contact the libav-devel mailing list.\n");
+ "of this file to ftp://upload.ffmpeg.org/MPlayer/incoming/ "
+ "and contact the ffmpeg-devel mailing list.\n");
va_end(argument_list);
}
int ff_thread_get_buffer(AVCodecContext *avctx, AVFrame *f)
{
f->owner = avctx;
+
+ ff_init_buffer_info(avctx, f);
+
return avctx->get_buffer(avctx, f);
}
enum AVMediaType avcodec_get_type(enum CodecID codec_id)
{
+ AVCodec *c= avcodec_find_decoder(codec_id);
+ if(!c)
+ c= avcodec_find_encoder(codec_id);
+ if(c)
+ return c->type;
+
if (codec_id <= CODEC_ID_NONE)
return AVMEDIA_TYPE_UNKNOWN;
else if (codec_id < CODEC_ID_FIRST_AUDIO)
/*
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * Libav is distributed in the hope that it will be useful,
+ * FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#define AVCODEC_VERSION_H
#define LIBAVCODEC_VERSION_MAJOR 53
-#define LIBAVCODEC_VERSION_MINOR 22
+#define LIBAVCODEC_VERSION_MINOR 37
#define LIBAVCODEC_VERSION_MICRO 0
#define LIBAVCODEC_VERSION_INT AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \
#ifndef FF_API_INTERNAL_CONTEXT
#define FF_API_INTERNAL_CONTEXT (LIBAVCODEC_VERSION_MAJOR < 54)
#endif
+ #ifndef FF_API_TIFFENC_COMPLEVEL
+ #define FF_API_TIFFENC_COMPLEVEL (LIBAVCODEC_VERSION_MAJOR < 54)
+ #endif
+
#endif /* AVCODEC_VERSION_H */
/*
* copyright (c) 2006 Michael Niedermayer <michaelni@gmx.at>
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * Libav is distributed in the hope that it will be useful,
+ * FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
*/
#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, \
*
* @{
*/
+#ifndef FF_API_OLD_EVAL_NAMES
+#define FF_API_OLD_EVAL_NAMES (LIBAVUTIL_VERSION_MAJOR < 52)
+#endif
#ifndef FF_API_GET_BITS_PER_SAMPLE_FMT
#define FF_API_GET_BITS_PER_SAMPLE_FMT (LIBAVUTIL_VERSION_MAJOR < 52)
#endif
AVMEDIA_TYPE_NB
};
+/**
+ * Return a string describing the media_type enum, NULL if media_type
+ * is unknown.
+ */
+const char *av_get_media_type_string(enum AVMediaType media_type);
+
/**
* @defgroup lavu_const Constants
* @{
*/
enum AVPictureType {
- AV_PICTURE_TYPE_I = 1, ///< Intra
+ AV_PICTURE_TYPE_NONE = 0, ///< Undefined
+ AV_PICTURE_TYPE_I, ///< Intra
AV_PICTURE_TYPE_P, ///< Predicted
AV_PICTURE_TYPE_B, ///< Bi-dir predicted
AV_PICTURE_TYPE_S, ///< S(GMC)-VOP MPEG4
* @}
*/
+/**
+ * Return x default pointer in case p is NULL.
+ */
+static inline const void *av_x_if_null(const void *p, const void *x)
+{
+ return p ? p : x;
+}
+
#include "common.h"
#include "error.h"
+#include "mathematics.h"
+#include "rational.h"
+#include "intfloat_readwrite.h"
+#include "log.h"
+#include "pixfmt.h"
/**
* @}
* pixel format descriptor
* Copyright (c) 2009 Michael Niedermayer <michaelni@gmx.at>
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * Libav is distributed in the hope that it will be useful,
+ * FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
{1,0,1,0,7}, /* U */
{2,0,1,0,7}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_YUYV422] = {
.name = "yuyv422",
{0,2,2,0,7}, /* G */
{0,2,3,0,7}, /* B */
},
+ .flags = PIX_FMT_RGB,
},
[PIX_FMT_BGR24] = {
.name = "bgr24",
{0,2,2,0,7}, /* G */
{0,2,3,0,7}, /* R */
},
+ .flags = PIX_FMT_RGB,
},
[PIX_FMT_YUV422P] = {
.name = "yuv422p",
{1,0,1,0,7}, /* U */
{2,0,1,0,7}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_YUV444P] = {
.name = "yuv444p",
{1,0,1,0,7}, /* U */
{2,0,1,0,7}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_YUV410P] = {
.name = "yuv410p",
{1,0,1,0,7}, /* U */
{2,0,1,0,7}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_YUV411P] = {
.name = "yuv411p",
{1,0,1,0,7}, /* U */
{2,0,1,0,7}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_GRAY8] = {
.name = "gray",
.comp = {
{0,0,1,0,7}, /* Y */
},
- .flags = PIX_FMT_PAL,
},
[PIX_FMT_MONOWHITE] = {
.name = "monow",
{1,0,1,0,7}, /* U */
{2,0,1,0,7}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_YUVJ422P] = {
.name = "yuvj422p",
{1,0,1,0,7}, /* U */
{2,0,1,0,7}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_YUVJ444P] = {
.name = "yuvj444p",
{1,0,1,0,7}, /* U */
{2,0,1,0,7}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_XVMC_MPEG2_MC] = {
.name = "xvmcmc",
{0,0,1,3,2}, /* G */
{0,0,1,0,2}, /* R */
},
- .flags = PIX_FMT_PAL,
+ .flags = PIX_FMT_PAL | PIX_FMT_RGB,
},
[PIX_FMT_BGR4] = {
.name = "bgr4",
{0,3,2,0,1}, /* G */
{0,3,4,0,0}, /* R */
},
- .flags = PIX_FMT_BITSTREAM,
+ .flags = PIX_FMT_BITSTREAM | PIX_FMT_RGB,
},
[PIX_FMT_BGR4_BYTE] = {
.name = "bgr4_byte",
{0,0,1,1,1}, /* G */
{0,0,1,0,0}, /* R */
},
- .flags = PIX_FMT_PAL,
+ .flags = PIX_FMT_PAL | PIX_FMT_RGB,
},
[PIX_FMT_RGB8] = {
.name = "rgb8",
{0,0,1,3,2}, /* G */
{0,0,1,0,2}, /* B */
},
- .flags = PIX_FMT_PAL,
+ .flags = PIX_FMT_PAL | PIX_FMT_RGB,
},
[PIX_FMT_RGB4] = {
.name = "rgb4",
{0,3,2,0,1}, /* G */
{0,3,4,0,0}, /* B */
},
- .flags = PIX_FMT_BITSTREAM,
+ .flags = PIX_FMT_BITSTREAM | PIX_FMT_RGB,
},
[PIX_FMT_RGB4_BYTE] = {
.name = "rgb4_byte",
{0,0,1,1,1}, /* G */
{0,0,1,0,0}, /* B */
},
- .flags = PIX_FMT_PAL,
+ .flags = PIX_FMT_PAL | PIX_FMT_RGB,
},
[PIX_FMT_NV12] = {
.name = "nv12",
{1,1,1,0,7}, /* U */
{1,1,2,0,7}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_NV21] = {
.name = "nv21",
{1,1,1,0,7}, /* V */
{1,1,2,0,7}, /* U */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_ARGB] = {
.name = "argb",
{0,3,3,0,7}, /* G */
{0,3,4,0,7}, /* B */
},
+ .flags = PIX_FMT_RGB,
},
[PIX_FMT_RGBA] = {
.name = "rgba",
{0,3,3,0,7}, /* B */
{0,3,4,0,7}, /* A */
},
+ .flags = PIX_FMT_RGB,
},
[PIX_FMT_ABGR] = {
.name = "abgr",
{0,3,3,0,7}, /* G */
{0,3,4,0,7}, /* R */
},
+ .flags = PIX_FMT_RGB,
},
[PIX_FMT_BGRA] = {
.name = "bgra",
{0,3,3,0,7}, /* R */
{0,3,4,0,7}, /* A */
},
+ .flags = PIX_FMT_RGB,
},
[PIX_FMT_GRAY16BE] = {
.name = "gray16be",
{1,0,1,0,7}, /* U */
{2,0,1,0,7}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_YUVJ440P] = {
.name = "yuvj440p",
{1,0,1,0,7}, /* U */
{2,0,1,0,7}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_YUVA420P] = {
.name = "yuva420p",
{2,0,1,0,7}, /* V */
{3,0,1,0,7}, /* A */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_VDPAU_H264] = {
.name = "vdpau_h264",
{0,5,3,0,15}, /* G */
{0,5,5,0,15}, /* B */
},
- .flags = PIX_FMT_BE,
+ .flags = PIX_FMT_RGB | PIX_FMT_BE,
},
[PIX_FMT_RGB48LE] = {
.name = "rgb48le",
{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,
{0,1,1,5,5}, /* G */
{0,1,1,0,4}, /* B */
},
- .flags = PIX_FMT_BE,
+ .flags = PIX_FMT_BE | PIX_FMT_RGB,
},
[PIX_FMT_RGB565LE] = {
.name = "rgb565le",
{0,1,1,5,5}, /* G */
{0,1,1,0,4}, /* B */
},
+ .flags = PIX_FMT_RGB,
},
[PIX_FMT_RGB555BE] = {
.name = "rgb555be",
{0,1,1,5,4}, /* G */
{0,1,1,0,4}, /* B */
},
- .flags = PIX_FMT_BE,
+ .flags = PIX_FMT_BE | PIX_FMT_RGB,
},
[PIX_FMT_RGB555LE] = {
.name = "rgb555le",
{0,1,1,5,4}, /* G */
{0,1,1,0,4}, /* B */
},
+ .flags = PIX_FMT_RGB,
},
[PIX_FMT_RGB444BE] = {
.name = "rgb444be",
{0,1,1,4,3}, /* G */
{0,1,1,0,3}, /* B */
},
- .flags = PIX_FMT_BE,
+ .flags = PIX_FMT_BE | PIX_FMT_RGB,
},
[PIX_FMT_RGB444LE] = {
.name = "rgb444le",
{0,1,1,4,3}, /* G */
{0,1,1,0,3}, /* B */
},
+ .flags = PIX_FMT_RGB,
},
[PIX_FMT_BGR48BE] = {
.name = "bgr48be",
{0,5,3,0,15}, /* G */
{0,5,5,0,15}, /* R */
},
- .flags = PIX_FMT_BE,
+ .flags = PIX_FMT_BE | PIX_FMT_RGB,
},
[PIX_FMT_BGR48LE] = {
.name = "bgr48le",
{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,
{0,1,1,5,5}, /* G */
{0,1,1,0,4}, /* R */
},
- .flags = PIX_FMT_BE,
+ .flags = PIX_FMT_BE | PIX_FMT_RGB,
},
[PIX_FMT_BGR565LE] = {
.name = "bgr565le",
{0,1,1,5,5}, /* G */
{0,1,1,0,4}, /* R */
},
+ .flags = PIX_FMT_RGB,
},
[PIX_FMT_BGR555BE] = {
.name = "bgr555be",
{0,1,1,5,4}, /* G */
{0,1,1,0,4}, /* R */
},
- .flags = PIX_FMT_BE,
+ .flags = PIX_FMT_BE | PIX_FMT_RGB,
},
[PIX_FMT_BGR555LE] = {
.name = "bgr555le",
{0,1,1,5,4}, /* G */
{0,1,1,0,4}, /* R */
},
+ .flags = PIX_FMT_RGB,
},
[PIX_FMT_BGR444BE] = {
.name = "bgr444be",
{0,1,1,4,3}, /* G */
{0,1,1,0,3}, /* R */
},
- .flags = PIX_FMT_BE,
+ .flags = PIX_FMT_BE | PIX_FMT_RGB,
},
[PIX_FMT_BGR444LE] = {
.name = "bgr444le",
{0,1,1,4,3}, /* G */
{0,1,1,0,3}, /* R */
},
+ .flags = PIX_FMT_RGB,
},
[PIX_FMT_VAAPI_MOCO] = {
.name = "vaapi_moco",
{1,1,1,0,8}, /* U */
{2,1,1,0,8}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_YUV420P9BE] = {
.name = "yuv420p9be",
{1,1,1,0,8}, /* U */
{2,1,1,0,8}, /* V */
},
- .flags = PIX_FMT_BE,
+ .flags = PIX_FMT_BE | PIX_FMT_PLANAR,
},
[PIX_FMT_YUV420P10LE] = {
.name = "yuv420p10le",
{1,1,1,0,9}, /* U */
{2,1,1,0,9}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_YUV420P10BE] = {
.name = "yuv420p10be",
{1,1,1,0,9}, /* U */
{2,1,1,0,9}, /* V */
},
- .flags = PIX_FMT_BE,
+ .flags = PIX_FMT_BE | PIX_FMT_PLANAR,
},
[PIX_FMT_YUV420P16LE] = {
.name = "yuv420p16le",
{1,1,1,0,15}, /* U */
{2,1,1,0,15}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_YUV420P16BE] = {
.name = "yuv420p16be",
{1,1,1,0,15}, /* U */
{2,1,1,0,15}, /* V */
},
- .flags = PIX_FMT_BE,
+ .flags = PIX_FMT_BE | PIX_FMT_PLANAR,
},
[PIX_FMT_YUV422P9LE] = {
.name = "yuv422p9le",
{1,1,1,0,8}, /* U */
{2,1,1,0,8}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_YUV422P9BE] = {
.name = "yuv422p9be",
{1,1,1,0,8}, /* U */
{2,1,1,0,8}, /* V */
},
- .flags = PIX_FMT_BE,
+ .flags = PIX_FMT_BE | PIX_FMT_PLANAR,
},
[PIX_FMT_YUV422P10LE] = {
.name = "yuv422p10le",
{1,1,1,0,9}, /* U */
{2,1,1,0,9}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_YUV422P10BE] = {
.name = "yuv422p10be",
{1,1,1,0,9}, /* U */
{2,1,1,0,9}, /* V */
},
- .flags = PIX_FMT_BE,
+ .flags = PIX_FMT_BE | PIX_FMT_PLANAR,
},
[PIX_FMT_YUV422P16LE] = {
.name = "yuv422p16le",
{1,1,1,0,15}, /* U */
{2,1,1,0,15}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_YUV422P16BE] = {
.name = "yuv422p16be",
{1,1,1,0,15}, /* U */
{2,1,1,0,15}, /* V */
},
- .flags = PIX_FMT_BE,
+ .flags = PIX_FMT_BE | PIX_FMT_PLANAR,
},
[PIX_FMT_YUV444P16LE] = {
.name = "yuv444p16le",
{1,1,1,0,15}, /* U */
{2,1,1,0,15}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_YUV444P16BE] = {
.name = "yuv444p16be",
{1,1,1,0,15}, /* U */
{2,1,1,0,15}, /* V */
},
- .flags = PIX_FMT_BE,
+ .flags = PIX_FMT_BE | PIX_FMT_PLANAR,
},
[PIX_FMT_YUV444P10LE] = {
.name = "yuv444p10le",
{1,1,1,0,9}, /* U */
{2,1,1,0,9}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_YUV444P10BE] = {
.name = "yuv444p10be",
{1,1,1,0,9}, /* U */
{2,1,1,0,9}, /* V */
},
- .flags = PIX_FMT_BE,
+ .flags = PIX_FMT_BE | PIX_FMT_PLANAR,
},
[PIX_FMT_YUV444P9LE] = {
.name = "yuv444p9le",
{1,1,1,0,8}, /* U */
{2,1,1,0,8}, /* V */
},
+ .flags = PIX_FMT_PLANAR,
},
[PIX_FMT_YUV444P9BE] = {
.name = "yuv444p9be",
{1,1,1,0,8}, /* U */
{2,1,1,0,8}, /* V */
},
- .flags = PIX_FMT_BE,
+ .flags = PIX_FMT_BE | PIX_FMT_PLANAR,
},
[PIX_FMT_DXVA2_VLD] = {
.name = "dxva2_vld",
.log2_chroma_h = 1,
.flags = PIX_FMT_HWACCEL,
},
- [PIX_FMT_Y400A] = {
- .name = "y400a",
+ [PIX_FMT_VDA_VLD] = {
+ .name = "vda_vld",
+ .log2_chroma_w = 1,
+ .log2_chroma_h = 1,
+ .flags = PIX_FMT_HWACCEL,
+ },
+ [PIX_FMT_GRAY8A] = {
+ .name = "gray8a",
.nb_components= 2,
.comp = {
{0,1,1,0,7}, /* Y */
{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,
},
};
* pixel format descriptor
* Copyright (c) 2009 Michael Niedermayer <michaelni@gmx.at>
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * Libav is distributed in the hope that it will be useful,
+ * FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#define PIX_FMT_PAL 2 ///< Pixel format has a palette in data[1], values are indexes in this palette.
#define PIX_FMT_BITSTREAM 4 ///< All values of a component are bit-wise packed end to end.
#define PIX_FMT_HWACCEL 8 ///< Pixel format is an HW accelerated format.
+ #define PIX_FMT_PLANAR 16 ///< At least one pixel component is not in the first data plane
+ #define PIX_FMT_RGB 32 ///< The pixel format contains RGB-like data (as opposed to YUV/grayscale)
/**
* The array of all the pixel format descriptors.
/*
* copyright (c) 2006 Michael Niedermayer <michaelni@gmx.at>
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * Libav is distributed in the hope that it will be useful,
+ * FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
PIX_FMT_RGB444BE, ///< packed RGB 4:4:4, 16bpp, (msb)4A 4R 4G 4B(lsb), big-endian, most significant bits to 0
PIX_FMT_BGR444LE, ///< packed BGR 4:4:4, 16bpp, (msb)4A 4B 4G 4R(lsb), little-endian, most significant bits to 1
PIX_FMT_BGR444BE, ///< packed BGR 4:4:4, 16bpp, (msb)4A 4B 4G 4R(lsb), big-endian, most significant bits to 1
- PIX_FMT_Y400A, ///< 8bit gray, 8bit alpha
+ PIX_FMT_GRAY8A, ///< 8bit gray, 8bit alpha
PIX_FMT_BGR48BE, ///< packed RGB 16:16:16, 48bpp, 16B, 16G, 16R, the 2-byte value for each R/G/B component is stored as big-endian
PIX_FMT_BGR48LE, ///< packed RGB 16:16:16, 48bpp, 16B, 16G, 16R, the 2-byte value for each R/G/B component is stored as little-endian
+
+ //the following 10 formats have the disadvantage of needing 1 format for each bit depth, thus
+ //If you want to support multiple bit depths, then using PIX_FMT_YUV420P16* with the bpp stored seperately
+ //is better
PIX_FMT_YUV420P9BE, ///< planar YUV 4:2:0, 13.5bpp, (1 Cr & Cb sample per 2x2 Y samples), big-endian
PIX_FMT_YUV420P9LE, ///< planar YUV 4:2:0, 13.5bpp, (1 Cr & Cb sample per 2x2 Y samples), little-endian
PIX_FMT_YUV420P10BE,///< planar YUV 4:2:0, 15bpp, (1 Cr & Cb sample per 2x2 Y samples), big-endian
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 */
/*
- * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
+ * Copyright (C) 2001-2011 Michael Niedermayer <michaelni@gmx.at>
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * Libav is distributed in the hope that it will be useful,
+ * FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "swscale.h"
#include "swscale_internal.h"
#include "rgb2rgb.h"
+#include "libavutil/avassert.h"
#include "libavutil/intreadwrite.h"
#include "libavutil/cpu.h"
#include "libavutil/avutil.h"
#include "libavutil/bswap.h"
#include "libavutil/pixdesc.h"
-#define DITHER1XBPP
#define RGB2YUV_SHIFT 15
#define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
DECLARE_ALIGNED(8, const uint8_t, ff_sws_pb_64)[8] =
{ 64, 64, 64, 64, 64, 64, 64, 64 };
+DECLARE_ALIGNED(8, const uint8_t, dithers)[8][8][8]={
+{
+ { 0, 1, 0, 1, 0, 1, 0, 1,},
+ { 1, 0, 1, 0, 1, 0, 1, 0,},
+ { 0, 1, 0, 1, 0, 1, 0, 1,},
+ { 1, 0, 1, 0, 1, 0, 1, 0,},
+ { 0, 1, 0, 1, 0, 1, 0, 1,},
+ { 1, 0, 1, 0, 1, 0, 1, 0,},
+ { 0, 1, 0, 1, 0, 1, 0, 1,},
+ { 1, 0, 1, 0, 1, 0, 1, 0,},
+},{
+ { 1, 2, 1, 2, 1, 2, 1, 2,},
+ { 3, 0, 3, 0, 3, 0, 3, 0,},
+ { 1, 2, 1, 2, 1, 2, 1, 2,},
+ { 3, 0, 3, 0, 3, 0, 3, 0,},
+ { 1, 2, 1, 2, 1, 2, 1, 2,},
+ { 3, 0, 3, 0, 3, 0, 3, 0,},
+ { 1, 2, 1, 2, 1, 2, 1, 2,},
+ { 3, 0, 3, 0, 3, 0, 3, 0,},
+},{
+ { 2, 4, 3, 5, 2, 4, 3, 5,},
+ { 6, 0, 7, 1, 6, 0, 7, 1,},
+ { 3, 5, 2, 4, 3, 5, 2, 4,},
+ { 7, 1, 6, 0, 7, 1, 6, 0,},
+ { 2, 4, 3, 5, 2, 4, 3, 5,},
+ { 6, 0, 7, 1, 6, 0, 7, 1,},
+ { 3, 5, 2, 4, 3, 5, 2, 4,},
+ { 7, 1, 6, 0, 7, 1, 6, 0,},
+},{
+ { 4, 8, 7, 11, 4, 8, 7, 11,},
+ { 12, 0, 15, 3, 12, 0, 15, 3,},
+ { 6, 10, 5, 9, 6, 10, 5, 9,},
+ { 14, 2, 13, 1, 14, 2, 13, 1,},
+ { 4, 8, 7, 11, 4, 8, 7, 11,},
+ { 12, 0, 15, 3, 12, 0, 15, 3,},
+ { 6, 10, 5, 9, 6, 10, 5, 9,},
+ { 14, 2, 13, 1, 14, 2, 13, 1,},
+},{
+ { 9, 17, 15, 23, 8, 16, 14, 22,},
+ { 25, 1, 31, 7, 24, 0, 30, 6,},
+ { 13, 21, 11, 19, 12, 20, 10, 18,},
+ { 29, 5, 27, 3, 28, 4, 26, 2,},
+ { 8, 16, 14, 22, 9, 17, 15, 23,},
+ { 24, 0, 30, 6, 25, 1, 31, 7,},
+ { 12, 20, 10, 18, 13, 21, 11, 19,},
+ { 28, 4, 26, 2, 29, 5, 27, 3,},
+},{
+ { 18, 34, 30, 46, 17, 33, 29, 45,},
+ { 50, 2, 62, 14, 49, 1, 61, 13,},
+ { 26, 42, 22, 38, 25, 41, 21, 37,},
+ { 58, 10, 54, 6, 57, 9, 53, 5,},
+ { 16, 32, 28, 44, 19, 35, 31, 47,},
+ { 48, 0, 60, 12, 51, 3, 63, 15,},
+ { 24, 40, 20, 36, 27, 43, 23, 39,},
+ { 56, 8, 52, 4, 59, 11, 55, 7,},
+},{
+ { 18, 34, 30, 46, 17, 33, 29, 45,},
+ { 50, 2, 62, 14, 49, 1, 61, 13,},
+ { 26, 42, 22, 38, 25, 41, 21, 37,},
+ { 58, 10, 54, 6, 57, 9, 53, 5,},
+ { 16, 32, 28, 44, 19, 35, 31, 47,},
+ { 48, 0, 60, 12, 51, 3, 63, 15,},
+ { 24, 40, 20, 36, 27, 43, 23, 39,},
+ { 56, 8, 52, 4, 59, 11, 55, 7,},
+},{
+ { 36, 68, 60, 92, 34, 66, 58, 90,},
+ { 100, 4,124, 28, 98, 2,122, 26,},
+ { 52, 84, 44, 76, 50, 82, 42, 74,},
+ { 116, 20,108, 12,114, 18,106, 10,},
+ { 32, 64, 56, 88, 38, 70, 62, 94,},
+ { 96, 0,120, 24,102, 6,126, 30,},
+ { 48, 80, 40, 72, 54, 86, 46, 78,},
+ { 112, 16,104, 8,118, 22,110, 14,},
+}};
+
+static const uint8_t flat64[8]={64,64,64,64,64,64,64,64};
+
+const uint16_t dither_scale[15][16]={
+{ 2, 3, 3, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,},
+{ 2, 3, 7, 7, 13, 13, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25,},
+{ 3, 3, 4, 15, 15, 29, 57, 57, 57, 113, 113, 113, 113, 113, 113, 113,},
+{ 3, 4, 4, 5, 31, 31, 61, 121, 241, 241, 241, 241, 481, 481, 481, 481,},
+{ 3, 4, 5, 5, 6, 63, 63, 125, 249, 497, 993, 993, 993, 993, 993, 1985,},
+{ 3, 5, 6, 6, 6, 7, 127, 127, 253, 505, 1009, 2017, 4033, 4033, 4033, 4033,},
+{ 3, 5, 6, 7, 7, 7, 8, 255, 255, 509, 1017, 2033, 4065, 8129,16257,16257,},
+{ 3, 5, 6, 8, 8, 8, 8, 9, 511, 511, 1021, 2041, 4081, 8161,16321,32641,},
+{ 3, 5, 7, 8, 9, 9, 9, 9, 10, 1023, 1023, 2045, 4089, 8177,16353,32705,},
+{ 3, 5, 7, 8, 10, 10, 10, 10, 10, 11, 2047, 2047, 4093, 8185,16369,32737,},
+{ 3, 5, 7, 8, 10, 11, 11, 11, 11, 11, 12, 4095, 4095, 8189,16377,32753,},
+{ 3, 5, 7, 9, 10, 12, 12, 12, 12, 12, 12, 13, 8191, 8191,16381,32761,},
+{ 3, 5, 7, 9, 10, 12, 13, 13, 13, 13, 13, 13, 14,16383,16383,32765,},
+{ 3, 5, 7, 9, 10, 12, 14, 14, 14, 14, 14, 14, 14, 15,32767,32767,},
+{ 3, 5, 7, 9, 11, 12, 14, 15, 15, 15, 15, 15, 15, 15, 16,65535,},
+};
+
#define output_pixel(pos, val, bias, signedness) \
if (big_endian) { \
AV_WB16(pos, bias + av_clip_ ## signedness ## 16(val >> shift)); \
int big_endian, int output_bits)
{
int i;
- int shift = 19 - output_bits;
+ int shift = 3;
+ av_assert0(output_bits == 16);
for (i = 0; i < dstW; i++) {
int val = src[i] + (1 << (shift - 1));
int big_endian, int output_bits)
{
int i;
- int shift = 15 + 16 - output_bits;
+ int shift = 15;
+ av_assert0(output_bits == 16);
for (i = 0; i < dstW; i++) {
- int val = 1 << (30-output_bits);
+ int val = 1 << (shift - 1);
int j;
/* range of val is [0,0x7FFFFFFF], so 31 bits, but with lanczos/spline
int shift = 11 + 16 - output_bits;
for (i = 0; i < dstW; i++) {
- int val = 1 << (26-output_bits);
+ int val = 1 << (shift - 1);
int j;
for (j = 0; j < filterSize; j++)
int i;
for (i = 0; i < (dstW >> 1); i++) {
- int Y1 = buf0[i * 2 ] << 1;
- int Y2 = buf0[i * 2 + 1] << 1;
+ int Y1 = (buf0[i * 2 ]+4)>>3;
+ int Y2 = (buf0[i * 2 + 1]+4)>>3;
output_pixel(&dest[i * 2 + 0], Y1);
output_pixel(&dest[i * 2 + 1], Y2);
for (i = 0; i < (dstW >> 1); i++) {
int Y1 = (buf0[i * 2] ) >> 2;
int Y2 = (buf0[i * 2 + 1]) >> 2;
- int U = (ubuf0[i] + ubuf1[i] + (-128 << 11)) >> 3;
- int V = (vbuf0[i] + vbuf1[i] + (-128 << 11)) >> 3;
+ int U = (ubuf0[i] + ubuf1[i] + (-128 << 12)) >> 3;
+ int V = (vbuf0[i] + vbuf1[i] + (-128 << 12)) >> 3;
int R, G, B;
Y1 -= c->yuv2rgb_y_offset;
#define r_b ((target == PIX_FMT_RGB24) ? r : b)
#define b_r ((target == PIX_FMT_RGB24) ? b : r)
+
dest[i * 6 + 0] = r_b[Y1];
dest[i * 6 + 1] = g[Y1];
dest[i * 6 + 2] = b_r[Y1];
for (i = 0; i < dstW; i++) {
int j;
- int Y = 0;
- int U = -128 << 19;
- int V = -128 << 19;
+ int Y = 1<<9;
+ int U = (1<<9)-(128 << 19);
+ int V = (1<<9)-(128 << 19);
int av_unused A;
int R, G, B;
U >>= 10;
V >>= 10;
if (hasAlpha) {
- A = 1 << 21;
+ A = 1 << 18;
for (j = 0; j < lumFilterSize; j++) {
A += alpSrc[j][i] * lumFilter[j];
}
dest[1] = B >> 22;
dest[2] = G >> 22;
dest[3] = R >> 22;
- dest += 4;
break;
case PIX_FMT_BGR24:
dest[0] = B >> 22;
#undef input_pixel
#define rgb48funcs(pattern, BE_LE, origin) \
-static void pattern ## 48 ## BE_LE ## ToY_c(uint8_t *_dst, const uint8_t *_src, \
+static void pattern ## 48 ## BE_LE ## ToY_c(uint8_t *_dst, const uint8_t *_src, const uint8_t *unused0, const uint8_t *unused1,\
int width, uint32_t *unused) \
{ \
const uint16_t *src = (const uint16_t *) _src; \
} \
\
static void pattern ## 48 ## BE_LE ## ToUV_c(uint8_t *_dstU, uint8_t *_dstV, \
- const uint8_t *_src1, const uint8_t *_src2, \
+ const uint8_t *unused0, const uint8_t *_src1, const uint8_t *_src2, \
int width, uint32_t *unused) \
{ \
const uint16_t *src1 = (const uint16_t *) _src1, \
} \
\
static void pattern ## 48 ## BE_LE ## ToUV_half_c(uint8_t *_dstU, uint8_t *_dstV, \
- const uint8_t *_src1, const uint8_t *_src2, \
+ const uint8_t *unused0, const uint8_t *_src1, const uint8_t *_src2, \
int width, uint32_t *unused) \
{ \
const uint16_t *src1 = (const uint16_t *) _src1, \
(isBE(origin) ? AV_RB16(&src[(i)*2]) : AV_RL16(&src[(i)*2])))
static av_always_inline void
-rgb16_32ToY_c_template(uint8_t *dst, const uint8_t *src,
+rgb16_32ToY_c_template(int16_t *dst, const uint8_t *src,
int width, enum PixelFormat origin,
int shr, int shg, int shb, int shp,
int maskr, int maskg, int maskb,
int rsh, int gsh, int bsh, int S)
{
const int ry = RY << rsh, gy = GY << gsh, by = BY << bsh,
- rnd = 33 << (S - 1);
+ rnd = (32<<((S)-1)) + (1<<(S-7));
int i;
for (i = 0; i < width; i++) {
int g = (px & maskg) >> shg;
int r = (px & maskr) >> shr;
- dst[i] = (ry * r + gy * g + by * b + rnd) >> S;
+ dst[i] = (ry * r + gy * g + by * b + rnd) >> ((S)-6);
}
}
static av_always_inline void
-rgb16_32ToUV_c_template(uint8_t *dstU, uint8_t *dstV,
+rgb16_32ToUV_c_template(int16_t *dstU, int16_t *dstV,
const uint8_t *src, int width,
enum PixelFormat origin,
int shr, int shg, int shb, int shp,
{
const int ru = RU << rsh, gu = GU << gsh, bu = BU << bsh,
rv = RV << rsh, gv = GV << gsh, bv = BV << bsh,
- rnd = 257 << (S - 1);
+ rnd = (256<<((S)-1)) + (1<<(S-7));
int i;
for (i = 0; i < width; i++) {
int g = (px & maskg) >> shg;
int r = (px & maskr) >> shr;
- dstU[i] = (ru * r + gu * g + bu * b + rnd) >> S;
- dstV[i] = (rv * r + gv * g + bv * b + rnd) >> S;
+ dstU[i] = (ru * r + gu * g + bu * b + rnd) >> ((S)-6);
+ dstV[i] = (rv * r + gv * g + bv * b + rnd) >> ((S)-6);
}
}
static av_always_inline void
-rgb16_32ToUV_half_c_template(uint8_t *dstU, uint8_t *dstV,
+rgb16_32ToUV_half_c_template(int16_t *dstU, int16_t *dstV,
const uint8_t *src, int width,
enum PixelFormat origin,
int shr, int shg, int shb, int shp,
{
const int ru = RU << rsh, gu = GU << gsh, bu = BU << bsh,
rv = RV << rsh, gv = GV << gsh, bv = BV << bsh,
- rnd = 257 << S, maskgx = ~(maskr | maskb);
+ rnd = (256U<<(S)) + (1<<(S-6)), maskgx = ~(maskr | maskb);
int i;
maskr |= maskr << 1; maskb |= maskb << 1; maskg |= maskg << 1;
}
r = (rb & maskr) >> shr;
- dstU[i] = (ru * r + gu * g + bu * b + rnd) >> (S + 1);
- dstV[i] = (rv * r + gv * g + bv * b + rnd) >> (S + 1);
+ dstU[i] = (ru * r + gu * g + bu * b + (unsigned)rnd) >> ((S)-6+1);
+ dstV[i] = (rv * r + gv * g + bv * b + (unsigned)rnd) >> ((S)-6+1);
}
}
#define rgb16_32_wrapper(fmt, name, shr, shg, shb, shp, maskr, \
maskg, maskb, rsh, gsh, bsh, S) \
-static void name ## ToY_c(uint8_t *dst, const uint8_t *src, \
+static void name ## ToY_c(uint8_t *dst, const uint8_t *src, const uint8_t *unused1, const uint8_t *unused2, \
int width, uint32_t *unused) \
{ \
- rgb16_32ToY_c_template(dst, src, width, fmt, shr, shg, shb, shp, \
+ rgb16_32ToY_c_template((int16_t*)dst, src, width, fmt, \
+ shr, shg, shb, shp, \
maskr, maskg, maskb, rsh, gsh, bsh, S); \
} \
\
static void name ## ToUV_c(uint8_t *dstU, uint8_t *dstV, \
- const uint8_t *src, const uint8_t *dummy, \
+ const uint8_t *unused0, const uint8_t *src, const uint8_t *dummy, \
int width, uint32_t *unused) \
{ \
- rgb16_32ToUV_c_template(dstU, dstV, src, width, fmt, shr, shg, shb, shp, \
+ rgb16_32ToUV_c_template((int16_t*)dstU, (int16_t*)dstV, src, width, fmt, \
+ shr, shg, shb, shp, \
maskr, maskg, maskb, rsh, gsh, bsh, S); \
} \
\
static void name ## ToUV_half_c(uint8_t *dstU, uint8_t *dstV, \
- const uint8_t *src, const uint8_t *dummy, \
+ const uint8_t *unused0, const uint8_t *src, const uint8_t *dummy, \
int width, uint32_t *unused) \
{ \
- rgb16_32ToUV_half_c_template(dstU, dstV, src, width, fmt, shr, shg, shb, shp, \
+ rgb16_32ToUV_half_c_template((int16_t*)dstU, (int16_t*)dstV, src, width, fmt, \
+ shr, shg, shb, shp, \
maskr, maskg, maskb, rsh, gsh, bsh, S); \
}
rgb16_32_wrapper(PIX_FMT_RGB565BE, rgb16be, 0, 0, 0, 0, 0xF800, 0x07E0, 0x001F, 0, 5, 11, RGB2YUV_SHIFT+8);
rgb16_32_wrapper(PIX_FMT_RGB555BE, rgb15be, 0, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, 0, 5, 10, RGB2YUV_SHIFT+7);
-static void abgrToA_c(uint8_t *dst, const uint8_t *src, int width, uint32_t *unused)
+static void gbr24pToY_c(uint16_t *dst, const uint8_t *gsrc, const uint8_t *bsrc, const uint8_t *rsrc,
+ int width, uint32_t *unused)
+{
+ int i;
+ for (i = 0; i < width; i++) {
+ unsigned int g = gsrc[i];
+ unsigned int b = bsrc[i];
+ unsigned int r = rsrc[i];
+
+ dst[i] = (RY*r + GY*g + BY*b + (0x801<<(RGB2YUV_SHIFT-7))) >> (RGB2YUV_SHIFT-6);
+ }
+}
+
+static void gbr24pToUV_c(uint16_t *dstU, uint16_t *dstV,
+ const uint8_t *gsrc, const uint8_t *bsrc, const uint8_t *rsrc,
+ int width, enum PixelFormat origin)
+{
+ int i;
+ for (i = 0; i < width; i++) {
+ unsigned int g = gsrc[i];
+ unsigned int b = bsrc[i];
+ unsigned int r = rsrc[i];
+
+ 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 gbr24pToUV_half_c(uint16_t *dstU, uint16_t *dstV,
+ const uint8_t *gsrc, const uint8_t *bsrc, const uint8_t *rsrc,
+ int width, enum PixelFormat origin)
+{
+ int i;
+ for (i = 0; i < width; i++) {
+ unsigned int g = gsrc[2*i] + gsrc[2*i+1];
+ unsigned int b = bsrc[2*i] + bsrc[2*i+1];
+ unsigned int r = rsrc[2*i] + rsrc[2*i+1];
+
+ dstU[i] = (RU*r + GU*g + BU*b + (0x4001<<(RGB2YUV_SHIFT-6))) >> (RGB2YUV_SHIFT-6+1);
+ dstV[i] = (RV*r + GV*g + BV*b + (0x4001<<(RGB2YUV_SHIFT-6))) >> (RGB2YUV_SHIFT-6+1);
+ }
+}
+
+static void abgrToA_c(int16_t *dst, const uint8_t *src, const uint8_t *unused1, const uint8_t *unused2, int width, uint32_t *unused)
{
int i;
for (i=0; i<width; i++) {
- dst[i]= src[4*i];
+ dst[i]= src[4*i]<<6;
}
}
-static void rgbaToA_c(uint8_t *dst, const uint8_t *src, int width, uint32_t *unused)
+static void rgbaToA_c(int16_t *dst, const uint8_t *src, const uint8_t *unused1, const uint8_t *unused2, int width, uint32_t *unused)
{
int i;
for (i=0; i<width; i++) {
- dst[i]= src[4*i+3];
+ dst[i]= src[4*i+3]<<6;
}
}
-static void palToY_c(uint8_t *dst, const uint8_t *src, int width, uint32_t *pal)
+static void palToA_c(int16_t *dst, const uint8_t *src, const uint8_t *unused1, const uint8_t *unused2, int width, uint32_t *pal)
{
int i;
for (i=0; i<width; i++) {
int d= src[i];
- dst[i]= pal[d] & 0xFF;
+ dst[i]= (pal[d] >> 24)<<6;
}
}
-static void palToUV_c(uint8_t *dstU, uint8_t *dstV,
- const uint8_t *src1, const uint8_t *src2,
- int width, uint32_t *pal)
+static void palToY_c(int16_t *dst, const uint8_t *src, const uint8_t *unused1, const uint8_t *unused2, long width, uint32_t *pal)
+{
+ int i;
+ for (i=0; i<width; i++) {
+ int d= src[i];
+
+ dst[i]= (pal[d] & 0xFF)<<6;
+ }
+}
+
+static void palToUV_c(uint16_t *dstU, int16_t *dstV,
+ const uint8_t *unused0, const uint8_t *src1, const uint8_t *src2,
+ int width, uint32_t *pal)
{
int i;
assert(src1 == src2);
for (i=0; i<width; i++) {
int p= pal[src1[i]];
- dstU[i]= p>>8;
- dstV[i]= p>>16;
+ dstU[i]= (uint8_t)(p>> 8)<<6;
+ dstV[i]= (uint8_t)(p>>16)<<6;
}
}
-static void monowhite2Y_c(uint8_t *dst, const uint8_t *src,
- int width, uint32_t *unused)
+static void monowhite2Y_c(int16_t *dst, const uint8_t *src, const uint8_t *unused1, const uint8_t *unused2, int width, uint32_t *unused)
{
int i, j;
for (i=0; i<width/8; i++) {
int d= ~src[i];
for(j=0; j<8; j++)
- dst[8*i+j]= ((d>>(7-j))&1)*255;
+ dst[8*i+j]= ((d>>(7-j))&1)*16383;
+ }
+ if(width&7){
+ int d= ~src[i];
+ for(j=0; j<(width&7); j++)
+ dst[8*i+j]= ((d>>(7-j))&1)*16383;
}
}
-static void monoblack2Y_c(uint8_t *dst, const uint8_t *src,
- int width, uint32_t *unused)
+static void monoblack2Y_c(int16_t *dst, const uint8_t *src, const uint8_t *unused1, const uint8_t *unused2, int width, uint32_t *unused)
{
int i, j;
for (i=0; i<width/8; i++) {
int d= src[i];
for(j=0; j<8; j++)
- dst[8*i+j]= ((d>>(7-j))&1)*255;
+ dst[8*i+j]= ((d>>(7-j))&1)*16383;
+ }
+ if(width&7){
+ int d= src[i];
+ for(j=0; j<(width&7); j++)
+ dst[8*i+j]= ((d>>(7-j))&1)*16383;
}
}
//FIXME yuy2* can read up to 7 samples too much
-static void yuy2ToY_c(uint8_t *dst, const uint8_t *src, int width,
+static void yuy2ToY_c(uint8_t *dst, const uint8_t *src, const uint8_t *unused1, const uint8_t *unused2, int width,
uint32_t *unused)
{
int i;
dst[i]= src[2*i];
}
-static void yuy2ToUV_c(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1,
+static void yuy2ToUV_c(uint8_t *dstU, uint8_t *dstV, const uint8_t *unused0, const uint8_t *src1,
const uint8_t *src2, int width, uint32_t *unused)
{
int i;
assert(src1 == src2);
}
-static void bswap16Y_c(uint8_t *_dst, const uint8_t *_src, int width, uint32_t *unused)
+static void bswap16Y_c(uint8_t *_dst, const uint8_t *_src, const uint8_t *unused1, const uint8_t *unused2, int width, uint32_t *unused)
{
int i;
const uint16_t *src = (const uint16_t *) _src;
}
}
-static void bswap16UV_c(uint8_t *_dstU, uint8_t *_dstV, const uint8_t *_src1,
+static void bswap16UV_c(uint8_t *_dstU, uint8_t *_dstV, const uint8_t *unused0, const uint8_t *_src1,
const uint8_t *_src2, int width, uint32_t *unused)
{
int i;
/* This is almost identical to the previous, end exists only because
* yuy2ToY/UV)(dst, src+1, ...) would have 100% unaligned accesses. */
-static void uyvyToY_c(uint8_t *dst, const uint8_t *src, int width,
+static void uyvyToY_c(uint8_t *dst, const uint8_t *src, const uint8_t *unused1, const uint8_t *unused2, int width,
uint32_t *unused)
{
int i;
dst[i]= src[2*i+1];
}
-static void uyvyToUV_c(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1,
+static void uyvyToUV_c(uint8_t *dstU, uint8_t *dstV, const uint8_t *unused0, const uint8_t *src1,
const uint8_t *src2, int width, uint32_t *unused)
{
int i;
}
static void nv12ToUV_c(uint8_t *dstU, uint8_t *dstV,
- const uint8_t *src1, const uint8_t *src2,
+ const uint8_t *unused0, const uint8_t *src1, const uint8_t *src2,
int width, uint32_t *unused)
{
nvXXtoUV_c(dstU, dstV, src1, width);
}
static void nv21ToUV_c(uint8_t *dstU, uint8_t *dstV,
- const uint8_t *src1, const uint8_t *src2,
+ const uint8_t *unused0, const uint8_t *src1, const uint8_t *src2,
int width, uint32_t *unused)
{
nvXXtoUV_c(dstV, dstU, src1, width);
#define input_pixel(pos) (isBE(origin) ? AV_RB16(pos) : AV_RL16(pos))
-static void bgr24ToY_c(uint8_t *dst, const uint8_t *src,
+static void bgr24ToY_c(int16_t *dst, const uint8_t *src, const uint8_t *unused1, const uint8_t *unused2,
int width, uint32_t *unused)
{
int i;
int g= src[i*3+1];
int r= src[i*3+2];
- dst[i]= ((RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
+ dst[i]= ((RY*r + GY*g + BY*b + (32<<(RGB2YUV_SHIFT-1)) + (1<<(RGB2YUV_SHIFT-7)))>>(RGB2YUV_SHIFT-6));
}
}
-static void bgr24ToUV_c(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1,
+static void bgr24ToUV_c(int16_t *dstU, int16_t *dstV, const uint8_t *unused0, const uint8_t *src1,
const uint8_t *src2, int width, uint32_t *unused)
{
int i;
int g= src1[3*i + 1];
int r= src1[3*i + 2];
- dstU[i]= (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT;
- dstV[i]= (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT;
+ dstU[i]= (RU*r + GU*g + BU*b + (256<<(RGB2YUV_SHIFT-1)) + (1<<(RGB2YUV_SHIFT-7)))>>(RGB2YUV_SHIFT-6);
+ dstV[i]= (RV*r + GV*g + BV*b + (256<<(RGB2YUV_SHIFT-1)) + (1<<(RGB2YUV_SHIFT-7)))>>(RGB2YUV_SHIFT-6);
}
assert(src1 == src2);
}
-static void bgr24ToUV_half_c(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1,
+static void bgr24ToUV_half_c(int16_t *dstU, int16_t *dstV, const uint8_t *unused0, const uint8_t *src1,
const uint8_t *src2, int width, uint32_t *unused)
{
int i;
int g= src1[6*i + 1] + src1[6*i + 4];
int r= src1[6*i + 2] + src1[6*i + 5];
- 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 + (256<<RGB2YUV_SHIFT) + (1<<(RGB2YUV_SHIFT-6)))>>(RGB2YUV_SHIFT-5);
+ dstV[i]= (RV*r + GV*g + BV*b + (256<<RGB2YUV_SHIFT) + (1<<(RGB2YUV_SHIFT-6)))>>(RGB2YUV_SHIFT-5);
}
assert(src1 == src2);
}
-static void rgb24ToY_c(uint8_t *dst, const uint8_t *src, int width,
+static void rgb24ToY_c(int16_t *dst, const uint8_t *src, const uint8_t *unused1, const uint8_t *unused2, int width,
uint32_t *unused)
{
int i;
int g= src[i*3+1];
int b= src[i*3+2];
- dst[i]= ((RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
+ dst[i]= ((RY*r + GY*g + BY*b + (32<<(RGB2YUV_SHIFT-1)) + (1<<(RGB2YUV_SHIFT-7)))>>(RGB2YUV_SHIFT-6));
}
}
-static void rgb24ToUV_c(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1,
+static void rgb24ToUV_c(int16_t *dstU, int16_t *dstV, const uint8_t *unused0, const uint8_t *src1,
const uint8_t *src2, int width, uint32_t *unused)
{
int i;
int g= src1[3*i + 1];
int b= src1[3*i + 2];
- dstU[i]= (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT;
- dstV[i]= (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT;
+ dstU[i]= (RU*r + GU*g + BU*b + (256<<(RGB2YUV_SHIFT-1)) + (1<<(RGB2YUV_SHIFT-7)))>>(RGB2YUV_SHIFT-6);
+ dstV[i]= (RV*r + GV*g + BV*b + (256<<(RGB2YUV_SHIFT-1)) + (1<<(RGB2YUV_SHIFT-7)))>>(RGB2YUV_SHIFT-6);
}
}
-static void rgb24ToUV_half_c(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1,
- const uint8_t *src2, int width, uint32_t *unused)
+static void rgb24ToUV_half_c(int16_t *dstU, int16_t *dstV, const uint8_t *unused0, const uint8_t *src1,
+ const uint8_t *src2, int width, uint32_t *unused)
{
int i;
assert(src1==src2);
int g= src1[6*i + 1] + src1[6*i + 4];
int b= src1[6*i + 2] + src1[6*i + 5];
- 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 + (256<<RGB2YUV_SHIFT) + (1<<(RGB2YUV_SHIFT-6)))>>(RGB2YUV_SHIFT-5);
+ dstV[i]= (RV*r + GV*g + BV*b + (256<<RGB2YUV_SHIFT) + (1<<(RGB2YUV_SHIFT-6)))>>(RGB2YUV_SHIFT-5);
}
}
-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)
int bits = av_pix_fmt_descriptors[c->srcFormat].comp[0].depth_minus1;
int sh = bits - 4;
+ if((isAnyRGB(c->srcFormat) || c->srcFormat==PIX_FMT_PAL8) && av_pix_fmt_descriptors[c->srcFormat].comp[0].depth_minus1<15)
+ sh= 9;
+
for (i = 0; i < dstW; i++) {
int j;
int srcPos = filterPos[i];
const uint16_t *src = (const uint16_t *) _src;
int sh = av_pix_fmt_descriptors[c->srcFormat].comp[0].depth_minus1;
+ if(sh<15)
+ sh= isAnyRGB(c->srcFormat) || c->srcFormat==PIX_FMT_PAL8 ? 13 : av_pix_fmt_descriptors[c->srcFormat].comp[0].depth_minus1;
+
for (i = 0; i < dstW; i++) {
int j;
int srcPos = filterPos[i];
int i;
int32_t *dst = (int32_t *) _dst;
for (i = 0; i < width; i++)
- dst[i] = (dst[i]*14071 + (33561947<<4))>>14;
+ dst[i] = (dst[i]*(14071/4) + (33561947<<4)/4)>>12;
}
static void hyscale_fast_c(SwsContext *c, int16_t *dst, int dstWidth,
dst[i]= (src[xx]<<7) + (src[xx+1] - src[xx])*xalpha;
xpos+=xInc;
}
+ for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--)
+ dst[i] = src[srcW-1]*128;
}
// *** horizontal scale Y line to temp buffer
static av_always_inline void hyscale(SwsContext *c, int16_t *dst, int dstWidth,
- const uint8_t *src, const uint8_t *src2, const uint8_t *src3,
- int srcW, int xInc,
+ const uint8_t *src_in[4], int srcW, int xInc,
const int16_t *hLumFilter,
const int16_t *hLumFilterPos, int hLumFilterSize,
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) {
dst2[i]=(src2[xx]*(xalpha^127)+src2[xx+1]*xalpha);
xpos+=xInc;
}
+ for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) {
+ dst1[i] = src1[srcW-1]*128;
+ dst2[i] = src2[srcW-1]*128;
+ }
}
static av_always_inline void hcscale(SwsContext *c, int16_t *dst1, int16_t *dst2, int dstWidth,
- const uint8_t *src0, const uint8_t *src1, const uint8_t *src2,
+ const uint8_t *src_in[4],
int srcW, int xInc, const int16_t *hChrFilter,
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;
}
*yuv2packedX = yuv2bgr24_full_X_c;
break;
}
+ if(!*yuv2packedX)
+ goto YUV_PACKED;
} else {
+ YUV_PACKED:
switch (dstFormat) {
case PIX_FMT_GRAY16BE:
*yuv2packed1 = yuv2gray16BE_1_c;
int16_t *vChrFilterPos= c->vChrFilterPos;
int16_t *hLumFilterPos= c->hLumFilterPos;
int16_t *hChrFilterPos= c->hChrFilterPos;
- int16_t *vLumFilter= c->vLumFilter;
- int16_t *vChrFilter= c->vChrFilter;
int16_t *hLumFilter= c->hLumFilter;
int16_t *hChrFilter= c->hChrFilter;
int32_t *lumMmxFilter= c->lumMmxFilter;
const int chrSrcSliceH= -((-srcSliceH) >> c->chrSrcVSubSample);
int lastDstY;
uint32_t *pal=c->pal_yuv;
+ int should_dither= isNBPS(c->srcFormat) || is16BPS(c->srcFormat);
+
yuv2planar1_fn yuv2plane1 = c->yuv2plane1;
yuv2planarX_fn yuv2planeX = c->yuv2planeX;
yuv2interleavedX_fn yuv2nv12cX = c->yuv2nv12cX;
yuv2packed1_fn yuv2packed1 = c->yuv2packed1;
yuv2packed2_fn yuv2packed2 = c->yuv2packed2;
yuv2packedX_fn yuv2packedX = c->yuv2packedX;
- int should_dither = is9_OR_10BPS(c->srcFormat) || is16BPS(c->srcFormat);
/* vars which will change and which we need to store back in the context */
int dstY= c->dstY;
DEBUG_BUFFERS("vLumFilterSize: %d vLumBufSize: %d vChrFilterSize: %d vChrBufSize: %d\n",
vLumFilterSize, vLumBufSize, vChrFilterSize, vChrBufSize);
- if (dstStride[0]%8 !=0 || dstStride[1]%8 !=0 || dstStride[2]%8 !=0 || dstStride[3]%8 != 0) {
+ if (dstStride[0]%16 !=0 || dstStride[1]%16 !=0 || dstStride[2]%16 !=0 || dstStride[3]%16 != 0) {
static int warnedAlready=0; //FIXME move this into the context perhaps
if (flags & SWS_PRINT_INFO && !warnedAlready) {
av_log(c, AV_LOG_WARNING, "Warning: dstStride is not aligned!\n"
}
}
+ if ((int)dst[0]%16 || (int)dst[1]%16 || (int)dst[2]%16 || (int)src[0]%16 || (int)src[1]%16 || (int)src[2]%16
+ || dstStride[0]%16 || dstStride[1]%16 || dstStride[2]%16 || dstStride[3]%16
+ || srcStride[0]%16 || srcStride[1]%16 || srcStride[2]%16 || srcStride[3]%16
+ ) {
+ static int warnedAlready=0;
+ int cpu_flags = av_get_cpu_flags();
+ if (HAVE_MMX2 && (cpu_flags & AV_CPU_FLAG_SSE2) && !warnedAlready){
+ av_log(c, AV_LOG_WARNING, "Warning: data is not aligned! This can lead to a speedloss\n");
+ warnedAlready=1;
+ }
+ }
+
/* Note the user might start scaling the picture in the middle so this
will not get executed. This is not really intended but works
currently, so people might do it. */
dst[2] + dstStride[2] * chrDstY,
(CONFIG_SWSCALE_ALPHA && alpPixBuf) ? dst[3] + dstStride[3] * dstY : NULL,
};
+ int use_mmx_vfilter= c->use_mmx_vfilter;
const int firstLumSrcY= vLumFilterPos[dstY]; //First line needed as input
const int firstLumSrcY2= vLumFilterPos[FFMIN(dstY | ((1<<c->chrDstVSubSample) - 1), dstH-1)];
//Do horizontal scaling
while(lastInLumBuf < lastLumSrcY) {
- const uint8_t *src1= src[0]+(lastInLumBuf + 1 - srcSliceY)*srcStride[0];
- const uint8_t *src2= src[1]+(lastInLumBuf + 1 - srcSliceY)*srcStride[1];
- const uint8_t *src3= src[2]+(lastInLumBuf + 1 - srcSliceY)*srcStride[2];
- const uint8_t *src4= src[3]+(lastInLumBuf + 1 - srcSliceY)*srcStride[3];
+ const uint8_t *src1[4] = {
+ src[0] + (lastInLumBuf + 1 - srcSliceY) * srcStride[0],
+ src[1] + (lastInLumBuf + 1 - srcSliceY) * srcStride[1],
+ src[2] + (lastInLumBuf + 1 - srcSliceY) * srcStride[2],
+ src[3] + (lastInLumBuf + 1 - srcSliceY) * srcStride[3],
+ };
lumBufIndex++;
assert(lumBufIndex < 2*vLumBufSize);
assert(lastInLumBuf + 1 - srcSliceY < srcSliceH);
assert(lastInLumBuf + 1 - srcSliceY >= 0);
- hyscale(c, lumPixBuf[ lumBufIndex ], dstW, src1, src2, src3, srcW, lumXInc,
+ hyscale(c, lumPixBuf[ lumBufIndex ], dstW, src1, srcW, lumXInc,
hLumFilter, hLumFilterPos, hLumFilterSize,
formatConvBuffer,
pal, 0);
if (CONFIG_SWSCALE_ALPHA && alpPixBuf)
- hyscale(c, alpPixBuf[ lumBufIndex ], dstW, src4, NULL, NULL, srcW,
+ hyscale(c, alpPixBuf[ lumBufIndex ], dstW, src1, srcW,
lumXInc, hLumFilter, hLumFilterPos, hLumFilterSize,
formatConvBuffer,
pal, 1);
lumBufIndex, lastInLumBuf);
}
while(lastInChrBuf < lastChrSrcY) {
- const uint8_t *src0= src[0]+(lastInChrBuf + 1 - chrSrcSliceY)*srcStride[0];
- const uint8_t *src1= src[1]+(lastInChrBuf + 1 - chrSrcSliceY)*srcStride[1];
- const uint8_t *src2= src[2]+(lastInChrBuf + 1 - chrSrcSliceY)*srcStride[2];
+ const uint8_t *src1[4] = {
+ src[0] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[0],
+ src[1] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[1],
+ src[2] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[2],
+ src[3] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[3],
+ };
chrBufIndex++;
assert(chrBufIndex < 2*vChrBufSize);
assert(lastInChrBuf + 1 - chrSrcSliceY < (chrSrcSliceH));
if (c->needs_hcscale)
hcscale(c, chrUPixBuf[chrBufIndex], chrVPixBuf[chrBufIndex],
- chrDstW, src0, src1, src2, chrSrcW, chrXInc,
+ chrDstW, src1, chrSrcW, chrXInc,
hChrFilter, hChrFilterPos, hChrFilterSize,
formatConvBuffer, pal);
lastInChrBuf++;
// hmm looks like we can't use MMX here without overwriting this array's tail
find_c_packed_planar_out_funcs(c, &yuv2plane1, &yuv2planeX, &yuv2nv12cX,
&yuv2packed1, &yuv2packed2, &yuv2packedX);
+ use_mmx_vfilter= 0;
}
{
const int16_t **chrUSrcPtr= (const int16_t **) chrUPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
const int16_t **chrVSrcPtr= (const int16_t **) chrVPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
const int16_t **alpSrcPtr= (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? (const int16_t **) alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize : NULL;
+ int16_t *vLumFilter= c->vLumFilter;
+ int16_t *vChrFilter= c->vChrFilter;
+
if (isPlanarYUV(dstFormat) || dstFormat==PIX_FMT_GRAY8) { //YV12 like
const int chrSkipMask= (1<<c->chrDstVSubSample)-1;
+ vLumFilter += dstY * vLumFilterSize;
+ vChrFilter += chrDstY * vChrFilterSize;
+
+ av_assert0(use_mmx_vfilter != (
+ yuv2planeX == yuv2planeX_10BE_c
+ || yuv2planeX == yuv2planeX_10LE_c
+ || yuv2planeX == yuv2planeX_9BE_c
+ || yuv2planeX == yuv2planeX_9LE_c
+ || yuv2planeX == yuv2planeX_16BE_c
+ || yuv2planeX == yuv2planeX_16LE_c
+ || yuv2planeX == yuv2planeX_8_c) || !ARCH_X86);
+
+ if(use_mmx_vfilter){
+ vLumFilter= c->lumMmxFilter;
+ vChrFilter= c->chrMmxFilter;
+ }
+
if (vLumFilterSize == 1) {
yuv2plane1(lumSrcPtr[0], dest[0], dstW, c->lumDither8, 0);
} else {
- yuv2planeX(vLumFilter + dstY * vLumFilterSize, vLumFilterSize,
+ yuv2planeX(vLumFilter, vLumFilterSize,
lumSrcPtr, dest[0], dstW, c->lumDither8, 0);
}
if (!((dstY&chrSkipMask) || isGray(dstFormat))) {
if (yuv2nv12cX) {
- yuv2nv12cX(c, vChrFilter + chrDstY * vChrFilterSize, vChrFilterSize, chrUSrcPtr, chrVSrcPtr, dest[1], chrDstW);
+ yuv2nv12cX(c, vChrFilter, vChrFilterSize, chrUSrcPtr, chrVSrcPtr, dest[1], chrDstW);
} else if (vChrFilterSize == 1) {
yuv2plane1(chrUSrcPtr[0], dest[1], chrDstW, c->chrDither8, 0);
yuv2plane1(chrVSrcPtr[0], dest[2], chrDstW, c->chrDither8, 3);
} else {
- yuv2planeX(vChrFilter + chrDstY * vChrFilterSize, vChrFilterSize,
+ yuv2planeX(vChrFilter, vChrFilterSize,
chrUSrcPtr, dest[1], chrDstW, c->chrDither8, 0);
- yuv2planeX(vChrFilter + chrDstY * vChrFilterSize, vChrFilterSize,
- chrVSrcPtr, dest[2], chrDstW, c->chrDither8, 3);
+ yuv2planeX(vChrFilter, vChrFilterSize,
+ chrVSrcPtr, dest[2], chrDstW, c->chrDither8, use_mmx_vfilter ? (c->uv_offx2 >> 1) : 3);
}
}
if (CONFIG_SWSCALE_ALPHA && alpPixBuf){
+ if(use_mmx_vfilter){
+ vLumFilter= c->alpMmxFilter;
+ }
if (vLumFilterSize == 1) {
yuv2plane1(alpSrcPtr[0], dest[3], dstW, c->lumDither8, 0);
} else {
- yuv2planeX(vLumFilter + dstY * vLumFilterSize, vLumFilterSize,
+ yuv2planeX(vLumFilter, vLumFilterSize,
alpSrcPtr, dest[3], dstW, c->lumDither8, 0);
}
}
case PIX_FMT_PAL8 :
case PIX_FMT_BGR4_BYTE:
case PIX_FMT_RGB4_BYTE: c->chrToYV12 = palToUV_c; break;
+ case PIX_FMT_GBRP9LE:
+ case PIX_FMT_GBRP10LE:
+ case PIX_FMT_GBRP16LE: c->readChrPlanar = planar_rgb16le_to_uv; break;
+ case PIX_FMT_GBRP9BE:
+ case PIX_FMT_GBRP10BE:
+ case PIX_FMT_GBRP16BE: c->readChrPlanar = planar_rgb16be_to_uv; break;
+ case PIX_FMT_GBRP: c->readChrPlanar = planar_rgb_to_uv; break;
#if HAVE_BIGENDIAN
case PIX_FMT_YUV444P9LE:
case PIX_FMT_YUV422P9LE:
case PIX_FMT_YUV420P9LE:
case PIX_FMT_YUV422P10LE:
- case PIX_FMT_YUV444P10LE:
case PIX_FMT_YUV420P10LE:
+ case PIX_FMT_YUV444P10LE:
case PIX_FMT_YUV420P16LE:
case PIX_FMT_YUV422P16LE:
case PIX_FMT_YUV444P16LE: c->chrToYV12 = bswap16UV_c; break;
case PIX_FMT_RGB565BE: c->chrToYV12 = rgb16beToUV_half_c; break;
case PIX_FMT_RGB555LE: c->chrToYV12 = rgb15leToUV_half_c; break;
case PIX_FMT_RGB555BE: c->chrToYV12 = rgb15beToUV_half_c; break;
+ case PIX_FMT_GBR24P : c->chrToYV12 = gbr24pToUV_half_c; break;
}
} else {
switch(srcFormat) {
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;
}
}
c->lumToYV12 = NULL;
c->alpToYV12 = NULL;
switch (srcFormat) {
+ case PIX_FMT_GBRP9LE:
+ case PIX_FMT_GBRP10LE:
+ case PIX_FMT_GBRP16LE: c->readLumPlanar = planar_rgb16le_to_y; break;
+ case PIX_FMT_GBRP9BE:
+ case PIX_FMT_GBRP10BE:
+ case PIX_FMT_GBRP16BE: c->readLumPlanar = planar_rgb16be_to_y; break;
+ case PIX_FMT_GBRP: c->readLumPlanar = planar_rgb_to_y; break;
#if HAVE_BIGENDIAN
case PIX_FMT_YUV444P9LE:
case PIX_FMT_YUV422P9LE:
case PIX_FMT_YUV420P9LE:
- case PIX_FMT_YUV444P10LE:
case PIX_FMT_YUV422P10LE:
case PIX_FMT_YUV420P10LE:
+ case PIX_FMT_YUV444P10LE:
case PIX_FMT_YUV420P16LE:
case PIX_FMT_YUV422P16LE:
case PIX_FMT_YUV444P16LE:
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) {
case PIX_FMT_ABGR:
case PIX_FMT_ARGB: c->alpToYV12 = abgrToA_c; break;
case PIX_FMT_Y400A: c->alpToYV12 = uyvyToY_c; break;
+ case PIX_FMT_PAL8 : c->alpToYV12 = palToA_c; break;
}
}
+
if (c->srcBpc == 8) {
if (c->dstBpc <= 10) {
c->hyScale = c->hcScale = hScale8To15_c;
/*
- * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
+ * Copyright (C) 2001-2011 Michael Niedermayer <michaelni@gmx.at>
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * Libav is distributed in the hope that it will be useful,
+ * FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libavutil/avutil.h"
#include "libavutil/log.h"
#include "libavutil/pixfmt.h"
+ #include "libavutil/pixdesc.h"
#define STR(s) AV_TOSTRING(s) //AV_STRINGIFY is too long
#define MAX_FILTER_SIZE 256
+#define DITHER1XBPP
+
#if HAVE_BIGENDIAN
#define ALT32_CORR (-1)
#else
#define V_TEMP "11*8+4*4*256*2+32"
#define Y_TEMP "11*8+4*4*256*2+40"
#define ALP_MMX_FILTER_OFFSET "11*8+4*4*256*2+48"
-#define UV_OFF_PX "11*8+4*4*256*3+48"
-#define UV_OFF_BYTE "11*8+4*4*256*3+56"
+#define UV_OFF "11*8+4*4*256*3+48"
+#define UV_OFFx2 "11*8+4*4*256*3+56"
#define DITHER16 "11*8+4*4*256*3+64"
#define DITHER32 "11*8+4*4*256*3+80"
// alignment of these values is not necessary, but merely here
// to maintain the same offset across x8632 and x86-64. Once we
// use proper offset macros in the asm, they can be removed.
- DECLARE_ALIGNED(8, ptrdiff_t, uv_off_px); ///< offset (in pixels) between u and v planes
- DECLARE_ALIGNED(8, ptrdiff_t, uv_off_byte); ///< offset (in bytes) between u and v planes
+ DECLARE_ALIGNED(8, ptrdiff_t, uv_off); ///< offset (in pixels) between u and v planes
+ DECLARE_ALIGNED(8, ptrdiff_t, uv_offx2); ///< offset (in bytes) between u and v planes
DECLARE_ALIGNED(8, uint16_t, dither16)[8];
DECLARE_ALIGNED(8, uint32_t, dither32)[8];
#if HAVE_VIS
DECLARE_ALIGNED(8, uint64_t, sparc_coeffs)[10];
#endif
+ int use_mmx_vfilter;
/* function pointers for swScale() */
yuv2planar1_fn yuv2plane1;
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(),
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 \
)
- #define isGray16(x) ( \
- (x)==PIX_FMT_GRAY16BE \
- || (x)==PIX_FMT_GRAY16LE \
- )
+ #endif
+
#define isRGBinInt(x) ( \
(x)==PIX_FMT_RGB48BE \
|| (x)==PIX_FMT_RGB48LE \
+ || (x)==PIX_FMT_RGBA64BE \
+ || (x)==PIX_FMT_RGBA64LE \
|| (x)==PIX_FMT_RGB32 \
|| (x)==PIX_FMT_RGB32_1 \
|| (x)==PIX_FMT_RGB24 \
#define isBGRinInt(x) ( \
(x)==PIX_FMT_BGR48BE \
|| (x)==PIX_FMT_BGR48LE \
+ || (x)==PIX_FMT_BGRA64BE \
+ || (x)==PIX_FMT_BGRA64LE \
|| (x)==PIX_FMT_BGR32 \
|| (x)==PIX_FMT_BGR32_1 \
|| (x)==PIX_FMT_BGR24 \
|| (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];
+extern const uint8_t dithers[8][8][8];
+extern const uint16_t dither_scale[15][16];
+
extern const AVClass sws_context_class;
/*
- * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
+ * Copyright (C) 2001-2011 Michael Niedermayer <michaelni@gmx.at>
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * Libav is distributed in the hope that it will be useful,
+ * FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libavutil/bswap.h"
#include "libavutil/pixdesc.h"
-DECLARE_ALIGNED(8, const uint8_t, dither_8x8_1)[8][8] = {
- { 0, 1, 0, 1, 0, 1, 0, 1,},
- { 1, 0, 1, 0, 1, 0, 1, 0,},
- { 0, 1, 0, 1, 0, 1, 0, 1,},
- { 1, 0, 1, 0, 1, 0, 1, 0,},
- { 0, 1, 0, 1, 0, 1, 0, 1,},
- { 1, 0, 1, 0, 1, 0, 1, 0,},
- { 0, 1, 0, 1, 0, 1, 0, 1,},
- { 1, 0, 1, 0, 1, 0, 1, 0,},
-};
-DECLARE_ALIGNED(8, const uint8_t, dither_8x8_3)[8][8] = {
- { 1, 2, 1, 2, 1, 2, 1, 2,},
- { 3, 0, 3, 0, 3, 0, 3, 0,},
- { 1, 2, 1, 2, 1, 2, 1, 2,},
- { 3, 0, 3, 0, 3, 0, 3, 0,},
- { 1, 2, 1, 2, 1, 2, 1, 2,},
- { 3, 0, 3, 0, 3, 0, 3, 0,},
- { 1, 2, 1, 2, 1, 2, 1, 2,},
- { 3, 0, 3, 0, 3, 0, 3, 0,},
-};
-DECLARE_ALIGNED(8, const uint8_t, dither_8x8_64)[8][8] = {
- { 18, 34, 30, 46, 17, 33, 29, 45,},
- { 50, 2, 62, 14, 49, 1, 61, 13,},
- { 26, 42, 22, 38, 25, 41, 21, 37,},
- { 58, 10, 54, 6, 57, 9, 53, 5,},
- { 16, 32, 28, 44, 19, 35, 31, 47,},
- { 48, 0, 60, 12, 51, 3, 63, 15,},
- { 24, 40, 20, 36, 27, 43, 23, 39,},
- { 56, 8, 52, 4, 59, 11, 55, 7,},
-};
-extern const uint8_t dither_8x8_128[8][8];
-DECLARE_ALIGNED(8, const uint8_t, dither_8x8_256)[8][8] = {
- { 72, 136, 120, 184, 68, 132, 116, 180,},
- { 200, 8, 248, 56, 196, 4, 244, 52,},
- { 104, 168, 88, 152, 100, 164, 84, 148,},
- { 232, 40, 216, 24, 228, 36, 212, 20,},
- { 64, 128, 102, 176, 76, 140, 124, 188,},
- { 192, 0, 240, 48, 204, 12, 252, 60,},
- { 96, 160, 80, 144, 108, 172, 92, 156,},
- { 224, 32, 208, 16, 236, 44, 220, 28,},
-};
-
#define RGB2YUV_SHIFT 15
#define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
#define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
const uint8_t *srcPtr= src[0];
- if (srcFormat == PIX_FMT_Y400A) {
+ if (srcFormat == PIX_FMT_GRAY8A) {
switch (dstFormat) {
case PIX_FMT_RGB32 : conv = gray8aToPacked32; break;
case PIX_FMT_BGR32 : conv = gray8aToPacked32; break;
if (!conv)
av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
- sws_format_name(srcFormat), sws_format_name(dstFormat));
+ av_get_pix_fmt_name(srcFormat), av_get_pix_fmt_name(dstFormat));
else {
for (i=0; i<srcSliceH; i++) {
conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
return srcSliceH;
}
+static void gbr24ptopacked24(const uint8_t* src[], int srcStride[], uint8_t* dst, int dstStride, int srcSliceH, int width)
+{
+ int x, h, i;
+ for (h = 0; h < srcSliceH; h++) {
+ uint8_t *dest = dst + dstStride * h;
+ for (x = 0; x < width; x++) {
+ *dest++ = src[0][x];
+ *dest++ = src[1][x];
+ *dest++ = src[2][x];
+ }
+
+ for (i = 0; i < 3; i++)
+ src[i] += srcStride[i];
+ }
+}
+
+static void gbr24ptopacked32(const uint8_t* src[], int srcStride[], uint8_t* dst, int dstStride, int srcSliceH, int alpha_first, int width)
+{
+ int x, h, i;
+ for (h = 0; h < srcSliceH; h++) {
+ uint8_t *dest = dst + dstStride * h;
+
+ if (alpha_first) {
+ for (x = 0; x < width; x++) {
+ *dest++ = 0xff;
+ *dest++ = src[0][x];
+ *dest++ = src[1][x];
+ *dest++ = src[2][x];
+ }
+ } else {
+ for (x = 0; x < width; x++) {
+ *dest++ = src[0][x];
+ *dest++ = src[1][x];
+ *dest++ = src[2][x];
+ *dest++ = 0xff;
+ }
+ }
+
+ for (i = 0; i < 3; i++)
+ src[i] += srcStride[i];
+ }
+}
+
+static int planarRgbToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
+ int srcSliceH, uint8_t* dst[], int dstStride[])
+{
+ int alpha_first = 0;
+ if (c->srcFormat != PIX_FMT_GBR24P) {
+ av_log(c, AV_LOG_ERROR, "unsupported planar RGB conversion %s -> %s\n",
+ av_get_pix_fmt_name(c->srcFormat), av_get_pix_fmt_name(c->dstFormat));
+ return srcSliceH;
+ }
+
+ switch (c->dstFormat) {
+ case PIX_FMT_BGR24:
+ gbr24ptopacked24((const uint8_t* []) {src[1], src[0], src[2]}, (int []) {srcStride[1], srcStride[0], srcStride[2]},
+ dst[0] + srcSliceY * dstStride[0], dstStride[0], srcSliceH, c->srcW);
+ break;
+
+ case PIX_FMT_RGB24:
+ gbr24ptopacked24((const uint8_t* []) {src[2], src[0], src[1]}, (int []) {srcStride[2], srcStride[0], srcStride[1]},
+ dst[0] + srcSliceY * dstStride[0], dstStride[0], srcSliceH, c->srcW);
+ break;
+
+ case PIX_FMT_ARGB:
+ alpha_first = 1;
+ case PIX_FMT_RGBA:
+ gbr24ptopacked32((const uint8_t* []) {src[2], src[0], src[1]}, (int []) {srcStride[2], srcStride[0], srcStride[1]},
+ dst[0] + srcSliceY * dstStride[0], dstStride[0], srcSliceH, alpha_first, c->srcW);
+ break;
+
+ case PIX_FMT_ABGR:
+ alpha_first = 1;
+ case PIX_FMT_BGRA:
+ gbr24ptopacked32((const uint8_t* []) {src[1], src[0], src[2]}, (int []) {srcStride[1], srcStride[0], srcStride[2]},
+ dst[0] + srcSliceY * dstStride[0], dstStride[0], srcSliceH, alpha_first, c->srcW);
+ break;
+
+ default:
+ av_log(c, AV_LOG_ERROR, "unsupported planar RGB conversion %s -> %s\n",
+ av_get_pix_fmt_name(c->srcFormat), av_get_pix_fmt_name(c->dstFormat));
+ }
+
+ return srcSliceH;
+}
+
#define isRGBA32(x) ( \
(x) == PIX_FMT_ARGB \
|| (x) == PIX_FMT_RGBA \
if (!conv) {
av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
- sws_format_name(srcFormat), sws_format_name(dstFormat));
+ av_get_pix_fmt_name(srcFormat), av_get_pix_fmt_name(dstFormat));
} else {
const uint8_t *srcPtr= src[0];
uint8_t *dstPtr= dst[0];
return srcSliceH;
}
-#define clip9(x) av_clip_uintp2(x, 9)
-#define clip10(x) av_clip_uintp2(x, 10)
-#define DITHER_COPY(dst, dstStride, wfunc, src, srcStride, rfunc, dithers, shift, clip) \
- for (i = 0; i < height; i++) { \
- const uint8_t *dither = dithers[i & 7]; \
- for (j = 0; j < length - 7; j += 8) { \
- wfunc(&dst[j + 0], clip((rfunc(&src[j + 0]) + dither[0]) >> shift)); \
- wfunc(&dst[j + 1], clip((rfunc(&src[j + 1]) + dither[1]) >> shift)); \
- wfunc(&dst[j + 2], clip((rfunc(&src[j + 2]) + dither[2]) >> shift)); \
- wfunc(&dst[j + 3], clip((rfunc(&src[j + 3]) + dither[3]) >> shift)); \
- wfunc(&dst[j + 4], clip((rfunc(&src[j + 4]) + dither[4]) >> shift)); \
- wfunc(&dst[j + 5], clip((rfunc(&src[j + 5]) + dither[5]) >> shift)); \
- wfunc(&dst[j + 6], clip((rfunc(&src[j + 6]) + dither[6]) >> shift)); \
- wfunc(&dst[j + 7], clip((rfunc(&src[j + 7]) + dither[7]) >> shift)); \
- } \
- for (; j < length; j++) \
- wfunc(&dst[j], (rfunc(&src[j]) + dither[j & 7]) >> shift); \
- dst += dstStride; \
- src += srcStride; \
+#define DITHER_COPY(dst, dstStride, src, srcStride, bswap, dbswap)\
+ uint16_t scale= dither_scale[dst_depth-1][src_depth-1];\
+ int shift= src_depth-dst_depth + dither_scale[src_depth-2][dst_depth-1];\
+ for (i = 0; i < height; i++) {\
+ const uint8_t *dither= dithers[src_depth-9][i&7];\
+ for (j = 0; j < length-7; j+=8){\
+ dst[j+0] = dbswap((bswap(src[j+0]) + dither[0])*scale>>shift);\
+ dst[j+1] = dbswap((bswap(src[j+1]) + dither[1])*scale>>shift);\
+ dst[j+2] = dbswap((bswap(src[j+2]) + dither[2])*scale>>shift);\
+ dst[j+3] = dbswap((bswap(src[j+3]) + dither[3])*scale>>shift);\
+ dst[j+4] = dbswap((bswap(src[j+4]) + dither[4])*scale>>shift);\
+ dst[j+5] = dbswap((bswap(src[j+5]) + dither[5])*scale>>shift);\
+ dst[j+6] = dbswap((bswap(src[j+6]) + dither[6])*scale>>shift);\
+ dst[j+7] = dbswap((bswap(src[j+7]) + dither[7])*scale>>shift);\
+ }\
+ for (; j < length; j++)\
+ dst[j] = dbswap((bswap(src[j]) + dither[j&7])*scale>>shift);\
+ dst += dstStride;\
+ src += srcStride;\
}
+
static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
int srcSliceH, uint8_t* dst[], int dstStride[])
{
int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
const uint8_t *srcPtr= src[plane];
uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
+ int shiftonly= plane==1 || plane==2 || (!c->srcRange && plane==0);
if (!dst[plane]) continue;
// ignore palette for GRAY8
length*=2;
fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
} else {
- if(is9_OR_10BPS(c->srcFormat)) {
+ if(isNBPS(c->srcFormat) || isNBPS(c->dstFormat)
+ || (is16BPS(c->srcFormat) != is16BPS(c->dstFormat))
+ ) {
const int src_depth = av_pix_fmt_descriptors[c->srcFormat].comp[plane].depth_minus1+1;
const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1+1;
const uint16_t *srcPtr2 = (const uint16_t*)srcPtr;
+ uint16_t *dstPtr2 = (uint16_t*)dstPtr;
- if (is16BPS(c->dstFormat)) {
- uint16_t *dstPtr2 = (uint16_t*)dstPtr;
-#define COPY9_OR_10TO16(rfunc, wfunc) \
- for (i = 0; i < height; i++) { \
- for (j = 0; j < length; j++) { \
- int srcpx = rfunc(&srcPtr2[j]); \
- wfunc(&dstPtr2[j], (srcpx<<(16-src_depth)) | (srcpx>>(2*src_depth-16))); \
- } \
- dstPtr2 += dstStride[plane]/2; \
- srcPtr2 += srcStride[plane]/2; \
+ if (dst_depth == 8) {
+ if(isBE(c->srcFormat) == HAVE_BIGENDIAN){
+ DITHER_COPY(dstPtr, dstStride[plane], srcPtr2, srcStride[plane]/2, , )
+ } else {
+ DITHER_COPY(dstPtr, dstStride[plane], srcPtr2, srcStride[plane]/2, av_bswap16, )
}
- if (isBE(c->dstFormat)) {
- if (isBE(c->srcFormat)) {
- COPY9_OR_10TO16(AV_RB16, AV_WB16);
- } else {
- COPY9_OR_10TO16(AV_RL16, AV_WB16);
+ } else if (src_depth == 8) {
+ for (i = 0; i < height; i++) {
+ #define COPY816(w)\
+ if(shiftonly){\
+ for (j = 0; j < length; j++)\
+ w(&dstPtr2[j], srcPtr[j]<<(dst_depth-8));\
+ }else{\
+ for (j = 0; j < length; j++)\
+ w(&dstPtr2[j], (srcPtr[j]<<(dst_depth-8)) |\
+ (srcPtr[j]>>(2*8-dst_depth)));\
}
- } else {
- if (isBE(c->srcFormat)) {
- COPY9_OR_10TO16(AV_RB16, AV_WL16);
+ if(isBE(c->dstFormat)){
+ COPY816(AV_WB16)
} else {
- COPY9_OR_10TO16(AV_RL16, AV_WL16);
+ COPY816(AV_WL16)
}
+ dstPtr2 += dstStride[plane]/2;
+ srcPtr += srcStride[plane];
}
- } else if (is9_OR_10BPS(c->dstFormat)) {
- uint16_t *dstPtr2 = (uint16_t*)dstPtr;
-#define COPY9_OR_10TO9_OR_10(loop) \
- for (i = 0; i < height; i++) { \
- for (j = 0; j < length; j++) { \
- loop; \
- } \
- dstPtr2 += dstStride[plane]/2; \
- srcPtr2 += srcStride[plane]/2; \
- }
-#define COPY9_OR_10TO9_OR_10_2(rfunc, wfunc) \
- if (dst_depth > src_depth) { \
- COPY9_OR_10TO9_OR_10(int srcpx = rfunc(&srcPtr2[j]); \
- wfunc(&dstPtr2[j], (srcpx << 1) | (srcpx >> 9))); \
- } else if (dst_depth < src_depth) { \
- DITHER_COPY(dstPtr2, dstStride[plane]/2, wfunc, \
- srcPtr2, srcStride[plane]/2, rfunc, \
- dither_8x8_1, 1, clip9); \
- } else { \
- COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]))); \
- }
- if (isBE(c->dstFormat)) {
- if (isBE(c->srcFormat)) {
- COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WB16);
- } else {
- COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WB16);
- }
- } else {
- if (isBE(c->srcFormat)) {
- COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WL16);
+ } else if (src_depth <= dst_depth) {
+ for (i = 0; i < height; i++) {
+#define COPY_UP(r,w) \
+ if(shiftonly){\
+ for (j = 0; j < length; j++){ \
+ unsigned int v= r(&srcPtr2[j]);\
+ w(&dstPtr2[j], v<<(dst_depth-src_depth));\
+ }\
+ }else{\
+ for (j = 0; j < length; j++){ \
+ unsigned int v= r(&srcPtr2[j]);\
+ w(&dstPtr2[j], (v<<(dst_depth-src_depth)) | \
+ (v>>(2*src_depth-dst_depth)));\
+ }\
+ }
+ if(isBE(c->srcFormat)){
+ if(isBE(c->dstFormat)){
+ COPY_UP(AV_RB16, AV_WB16)
+ } else {
+ COPY_UP(AV_RB16, AV_WL16)
+ }
} else {
- COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WL16);
+ if(isBE(c->dstFormat)){
+ COPY_UP(AV_RL16, AV_WB16)
+ } else {
+ COPY_UP(AV_RL16, AV_WL16)
+ }
}
+ dstPtr2 += dstStride[plane]/2;
+ srcPtr2 += srcStride[plane]/2;
}
} else {
-#define W8(a, b) { *(a) = (b); }
-#define COPY9_OR_10TO8(rfunc) \
- if (src_depth == 9) { \
- DITHER_COPY(dstPtr, dstStride[plane], W8, \
- srcPtr2, srcStride[plane]/2, rfunc, \
- dither_8x8_1, 1, av_clip_uint8); \
- } else { \
- DITHER_COPY(dstPtr, dstStride[plane], W8, \
- srcPtr2, srcStride[plane]/2, rfunc, \
- dither_8x8_3, 2, av_clip_uint8); \
- }
- if (isBE(c->srcFormat)) {
- COPY9_OR_10TO8(AV_RB16);
- } else {
- COPY9_OR_10TO8(AV_RL16);
- }
- }
- } else if(is9_OR_10BPS(c->dstFormat)) {
- const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1+1;
- uint16_t *dstPtr2 = (uint16_t*)dstPtr;
-
- if (is16BPS(c->srcFormat)) {
- const uint16_t *srcPtr2 = (const uint16_t*)srcPtr;
-#define COPY16TO9_OR_10(rfunc, wfunc) \
- if (dst_depth == 9) { \
- DITHER_COPY(dstPtr2, dstStride[plane]/2, wfunc, \
- srcPtr2, srcStride[plane]/2, rfunc, \
- dither_8x8_128, 7, clip9); \
- } else { \
- DITHER_COPY(dstPtr2, dstStride[plane]/2, wfunc, \
- srcPtr2, srcStride[plane]/2, rfunc, \
- dither_8x8_64, 6, clip10); \
- }
- if (isBE(c->dstFormat)) {
- if (isBE(c->srcFormat)) {
- COPY16TO9_OR_10(AV_RB16, AV_WB16);
+ if(isBE(c->srcFormat) == HAVE_BIGENDIAN){
+ if(isBE(c->dstFormat) == HAVE_BIGENDIAN){
+ DITHER_COPY(dstPtr2, dstStride[plane]/2, srcPtr2, srcStride[plane]/2, , )
} else {
- COPY16TO9_OR_10(AV_RL16, AV_WB16);
+ DITHER_COPY(dstPtr2, dstStride[plane]/2, srcPtr2, srcStride[plane]/2, , av_bswap16)
}
- } else {
- if (isBE(c->srcFormat)) {
- COPY16TO9_OR_10(AV_RB16, AV_WL16);
+ }else{
+ if(isBE(c->dstFormat) == HAVE_BIGENDIAN){
+ DITHER_COPY(dstPtr2, dstStride[plane]/2, srcPtr2, srcStride[plane]/2, av_bswap16, )
} else {
- COPY16TO9_OR_10(AV_RL16, AV_WL16);
+ DITHER_COPY(dstPtr2, dstStride[plane]/2, srcPtr2, srcStride[plane]/2, av_bswap16, av_bswap16)
}
}
- } else /* 8bit */ {
-#define COPY8TO9_OR_10(wfunc) \
- for (i = 0; i < height; i++) { \
- for (j = 0; j < length; j++) { \
- const int srcpx = srcPtr[j]; \
- wfunc(&dstPtr2[j], (srcpx<<(dst_depth-8)) | (srcpx >> (16-dst_depth))); \
- } \
- dstPtr2 += dstStride[plane]/2; \
- srcPtr += srcStride[plane]; \
- }
- if (isBE(c->dstFormat)) {
- COPY8TO9_OR_10(AV_WB16);
- } else {
- COPY8TO9_OR_10(AV_WL16);
- }
- }
- } else if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
- const uint16_t *srcPtr2 = (const uint16_t*)srcPtr;
-#define COPY16TO8(rfunc) \
- DITHER_COPY(dstPtr, dstStride[plane], W8, \
- srcPtr2, srcStride[plane]/2, rfunc, \
- dither_8x8_256, 8, av_clip_uint8);
- if (isBE(c->srcFormat)) {
- COPY16TO8(AV_RB16);
- } else {
- COPY16TO8(AV_RL16);
- }
- } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
- for (i=0; i<height; i++) {
- for (j=0; j<length; j++) {
- dstPtr[ j<<1 ] = srcPtr[j];
- dstPtr[(j<<1)+1] = srcPtr[j];
- }
- srcPtr+= srcStride[plane];
- dstPtr+= dstStride[plane];
}
} else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
&& isBE(c->srcFormat) != isBE(c->dstFormat)) {
&& (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
c->swScale= rgbToRgbWrapper;
- if ((usePal(srcFormat) && (
- dstFormat == PIX_FMT_RGB32 ||
- dstFormat == PIX_FMT_RGB32_1 ||
- dstFormat == PIX_FMT_RGB24 ||
- dstFormat == PIX_FMT_BGR32 ||
- dstFormat == PIX_FMT_BGR32_1 ||
- dstFormat == PIX_FMT_BGR24)))
+#define isByteRGB(f) (\
+ f == PIX_FMT_RGB32 ||\
+ f == PIX_FMT_RGB32_1 ||\
+ f == PIX_FMT_RGB24 ||\
+ f == PIX_FMT_BGR32 ||\
+ f == PIX_FMT_BGR32_1 ||\
+ f == PIX_FMT_BGR24)
+
+ if (isAnyRGB(srcFormat) && isPlanar(srcFormat) && isByteRGB(dstFormat))
+ c->swScale= planarRgbToRgbWrapper;
+
+ if (usePal(srcFormat) && isByteRGB(dstFormat))
c->swScale= palToRgbWrapper;
if (srcFormat == PIX_FMT_YUV422P) {
{
if(!isALPHA(format))
src[3]=NULL;
- if(!isPlanar(format)) {
+ if (!isPlanar(format)) {
src[3]=src[2]=NULL;
if (!usePal(format))
}
}
-static int check_image_pointers(uint8_t *data[4], enum PixelFormat pix_fmt,
+static int check_image_pointers(const uint8_t * const data[4], enum PixelFormat pix_fmt,
const int linesizes[4])
{
const AVPixFmtDescriptor *desc = &av_pix_fmt_descriptors[pix_fmt];
av_log(c, AV_LOG_ERROR, "bad src image pointers\n");
return 0;
}
- if (!check_image_pointers(dst, c->dstFormat, dstStride)) {
+ if (!check_image_pointers((const uint8_t* const*)dst, c->dstFormat, dstStride)) {
av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");
return 0;
}
if (usePal(c->srcFormat)) {
for (i=0; i<256; i++) {
- int p, r, g, b,y,u,v;
+ int p, r, g, b, y, u, v, a = 0xff;
if(c->srcFormat == PIX_FMT_PAL8) {
p=((const uint32_t*)(srcSlice[1]))[i];
+ a= (p>>24)&0xFF;
r= (p>>16)&0xFF;
g= (p>> 8)&0xFF;
b= p &0xFF;
r= (i>>3 )*255;
g= ((i>>1)&3)*85;
b= (i&1 )*255;
- } else if(c->srcFormat == PIX_FMT_GRAY8 || c->srcFormat == PIX_FMT_Y400A) {
+ } else if(c->srcFormat == PIX_FMT_GRAY8 || c->srcFormat == PIX_FMT_GRAY8A) {
r = g = b = i;
} else {
assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
- c->pal_yuv[i]= y + (u<<8) + (v<<16);
+ c->pal_yuv[i]= y + (u<<8) + (v<<16) + (a<<24);
switch(c->dstFormat) {
case PIX_FMT_BGR32:
#if !HAVE_BIGENDIAN
case PIX_FMT_RGB24:
#endif
- c->pal_rgb[i]= r + (g<<8) + (b<<16);
+ c->pal_rgb[i]= r + (g<<8) + (b<<16) + (a<<24);
break;
case PIX_FMT_BGR32_1:
#if HAVE_BIGENDIAN
case PIX_FMT_BGR24:
#endif
- c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
+ c->pal_rgb[i]= a + (r<<8) + (g<<16) + (b<<24);
break;
case PIX_FMT_RGB32_1:
#if HAVE_BIGENDIAN
case PIX_FMT_RGB24:
#endif
- c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
+ c->pal_rgb[i]= a + (b<<8) + (g<<16) + (r<<24);
break;
case PIX_FMT_RGB32:
#if !HAVE_BIGENDIAN
case PIX_FMT_BGR24:
#endif
default:
- c->pal_rgb[i]= b + (g<<8) + (r<<16);
+ c->pal_rgb[i]= b + (g<<8) + (r<<16) + (a<<24);
}
}
}
int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
reset_ptr(src2, c->srcFormat);
- reset_ptr((const uint8_t**)dst2, c->dstFormat);
+ reset_ptr((void*)dst2, c->dstFormat);
/* reset slice direction at end of frame */
if (srcSliceY + srcSliceH == c->srcH)
dst2[3] += ( c->dstH -1)*dstStride[3];
reset_ptr(src2, c->srcFormat);
- reset_ptr((const uint8_t**)dst2, c->dstFormat);
+ reset_ptr((void*)dst2, c->dstFormat);
/* reset slice direction at end of frame */
if (!srcSliceY)
/*
* Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * Libav is distributed in the hope that it will be useful,
+ * FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
+#define _SVID_SOURCE //needed for MAP_ANONYMOUS
+#define _DARWIN_C_SOURCE // needed for MAP_ANON
#include <inttypes.h>
#include <string.h>
#include <math.h>
#include "libavutil/bswap.h"
#include "libavutil/opt.h"
#include "libavutil/pixdesc.h"
+#include "libavutil/avassert.h"
unsigned swscale_version(void)
{
const char *swscale_configuration(void)
{
- return LIBAV_CONFIGURATION;
+ return FFMPEG_CONFIGURATION;
}
const char *swscale_license(void)
{
#define LICENSE_PREFIX "libswscale license: "
- return LICENSE_PREFIX LIBAV_LICENSE + sizeof(LICENSE_PREFIX) - 1;
+ return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
}
#define RET 0xC3 //near return opcode for x86
[PIX_FMT_YUVA420P] = { 1 , 1 },
[PIX_FMT_RGB48BE] = { 1 , 1 },
[PIX_FMT_RGB48LE] = { 1 , 1 },
+ [PIX_FMT_RGBA64BE] = { 0 , 0 },
+ [PIX_FMT_RGBA64LE] = { 0 , 0 },
[PIX_FMT_RGB565BE] = { 1 , 1 },
[PIX_FMT_RGB565LE] = { 1 , 1 },
[PIX_FMT_RGB555BE] = { 1 , 1 },
[PIX_FMT_Y400A] = { 1 , 0 },
[PIX_FMT_BGR48BE] = { 1 , 1 },
[PIX_FMT_BGR48LE] = { 1 , 1 },
+ [PIX_FMT_BGRA64BE] = { 0 , 0 },
+ [PIX_FMT_BGRA64LE] = { 0 , 0 },
[PIX_FMT_YUV420P9BE] = { 1 , 1 },
[PIX_FMT_YUV420P9LE] = { 1 , 1 },
[PIX_FMT_YUV420P10BE] = { 1 , 1 },
[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)
extern const int32_t ff_yuv2rgb_coeffs[8][4];
+#if FF_API_SWS_FORMAT_NAME
const char *sws_format_name(enum PixelFormat format)
{
- if ((unsigned)format < PIX_FMT_NB && av_pix_fmt_descriptors[format].name)
- return av_pix_fmt_descriptors[format].name;
- else
- return "Unknown format";
+ return av_get_pix_fmt_name(format);
}
+#endif
static double getSplineCoeff(double a, double b, double c, double d, double dist)
{
int *srcRange, int **table, int *dstRange,
int *brightness, int *contrast, int *saturation)
{
- if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
+ if (!c || isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
*inv_table = c->srcColorspaceTable;
*table = c->dstColorspaceTable;
int sws_init_context(SwsContext *c, SwsFilter *srcFilter, SwsFilter *dstFilter)
{
- int i;
+ int i, j;
int usesVFilter, usesHFilter;
int unscaled;
SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};
int srcH= c->srcH;
int dstW= c->dstW;
int dstH= c->dstH;
- int dst_stride = FFALIGN(dstW * sizeof(int16_t) + 16, 16), dst_stride_px = dst_stride >> 1;
+ int dst_stride = FFALIGN(dstW * sizeof(int16_t)+66, 16);
int flags, cpu_flags;
enum PixelFormat srcFormat= c->srcFormat;
enum PixelFormat dstFormat= c->dstFormat;
unscaled = (srcW == dstW && srcH == dstH);
+ handle_jpeg(&srcFormat);
+ handle_jpeg(&dstFormat);
+
+ if(srcFormat!=c->srcFormat || dstFormat!=c->dstFormat){
+ av_log(c, AV_LOG_WARNING, "deprecated pixel format used, make sure you did set range correctly\n");
+ c->srcFormat= srcFormat;
+ c->dstFormat= dstFormat;
+ }
+
if (!sws_isSupportedInput(srcFormat)) {
- av_log(c, AV_LOG_ERROR, "%s is not supported as input pixel format\n", sws_format_name(srcFormat));
+ av_log(c, AV_LOG_ERROR, "%s is not supported as input pixel format\n", av_get_pix_fmt_name(srcFormat));
return AVERROR(EINVAL);
}
if (!sws_isSupportedOutput(dstFormat)) {
- av_log(c, AV_LOG_ERROR, "%s is not supported as output pixel format\n", sws_format_name(dstFormat));
+ av_log(c, AV_LOG_ERROR, "%s is not supported as output pixel format\n", av_get_pix_fmt_name(dstFormat));
return AVERROR(EINVAL);
}
getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
// reuse chroma for 2 pixels RGB/BGR unless user wants full chroma interpolation
- if (flags & SWS_FULL_CHR_H_INT &&
- dstFormat != PIX_FMT_RGBA &&
- dstFormat != PIX_FMT_ARGB &&
- dstFormat != PIX_FMT_BGRA &&
- dstFormat != PIX_FMT_ABGR &&
- dstFormat != PIX_FMT_RGB24 &&
- dstFormat != PIX_FMT_BGR24) {
- av_log(c, AV_LOG_ERROR,
- "full chroma interpolation for destination format '%s' not yet implemented\n",
- sws_format_name(dstFormat));
- flags &= ~SWS_FULL_CHR_H_INT;
- c->flags = flags;
- }
if (isAnyRGB(dstFormat) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;
// drop some chroma lines if the user wants it
if (c->swScale) {
if (flags&SWS_PRINT_INFO)
av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",
- sws_format_name(srcFormat), sws_format_name(dstFormat));
+ av_get_pix_fmt_name(srcFormat), av_get_pix_fmt_name(dstFormat));
return 0;
}
}
c->dstBpc = 1 + av_pix_fmt_descriptors[dstFormat].comp[0].depth_minus1;
if (c->dstBpc < 8)
c->dstBpc = 8;
+ if (isAnyRGB(srcFormat) || srcFormat == PIX_FMT_PAL8)
+ c->srcBpc = 16;
if (c->dstBpc == 16)
dst_stride <<= 1;
- FF_ALLOC_OR_GOTO(c, c->formatConvBuffer,
- (FFALIGN(srcW, 16) * 2 * FFALIGN(c->srcBpc, 8) >> 3) + 16,
- fail);
+ FF_ALLOC_OR_GOTO(c, c->formatConvBuffer, FFALIGN(srcW*2+78, 16) * 2, fail);
if (HAVE_MMX2 && cpu_flags & AV_CPU_FLAG_MMX2 && c->srcBpc == 8 && c->dstBpc <= 10) {
c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR)) {
if (flags&SWS_PRINT_INFO)
av_log(c, AV_LOG_INFO, "output width is not a multiple of 32 -> no MMX2 scaler\n");
}
- if (usesHFilter) c->canMMX2BeUsed=0;
+ if (usesHFilter || isNBPS(c->srcFormat) || is16BPS(c->srcFormat) || isAnyRGB(c->srcFormat)) c->canMMX2BeUsed=0;
}
else
c->canMMX2BeUsed=0;
c->chrXInc+= 20;
}
//we don't use the x86 asm scaler if MMX is available
- else if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) {
+ else if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX && c->dstBpc <= 10) {
c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;
c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
}
c->chrMmx2FilterCode = av_malloc(c->chrMmx2FilterCodeSize);
#endif
+#ifdef MAP_ANONYMOUS
+ if (c->lumMmx2FilterCode == MAP_FAILED || c->chrMmx2FilterCode == MAP_FAILED)
+#else
if (!c->lumMmx2FilterCode || !c->chrMmx2FilterCode)
+#endif
return AVERROR(ENOMEM);
FF_ALLOCZ_OR_GOTO(c, c->hLumFilter , (dstW /8+8)*sizeof(int16_t), fail);
FF_ALLOCZ_OR_GOTO(c, c->hChrFilter , (c->chrDstW /4+8)*sizeof(int16_t), fail);
FF_ALLOCZ_OR_GOTO(c, c->hLumFilterPos, (dstW /2/8+8)*sizeof(int32_t), fail);
FF_ALLOCZ_OR_GOTO(c, c->hChrFilterPos, (c->chrDstW/2/4+8)*sizeof(int32_t), fail);
- initMMX2HScaler( dstW, c->lumXInc, c->lumMmx2FilterCode, c->hLumFilter, c->hLumFilterPos, 8);
- initMMX2HScaler(c->chrDstW, c->chrXInc, c->chrMmx2FilterCode, c->hChrFilter, c->hChrFilterPos, 4);
+ initMMX2HScaler( dstW, c->lumXInc, c->lumMmx2FilterCode, c->hLumFilter, (uint32_t*)c->hLumFilterPos, 8);
+ initMMX2HScaler(c->chrDstW, c->chrXInc, c->chrMmx2FilterCode, c->hChrFilter, (uint32_t*)c->hChrFilterPos, 4);
#ifdef MAP_ANONYMOUS
mprotect(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize, PROT_EXEC | PROT_READ);
c->vLumBufSize= c->vLumFilterSize;
c->vChrBufSize= c->vChrFilterSize;
for (i=0; i<dstH; i++) {
- int chrI= i*c->chrDstH / dstH;
+ int chrI= (int64_t)i*c->chrDstH / dstH;
int nextSlice= FFMAX(c->vLumFilterPos[i ] + c->vLumFilterSize - 1,
((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));
FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf[i+c->vLumBufSize], dst_stride+16, fail);
c->lumPixBuf[i] = c->lumPixBuf[i+c->vLumBufSize];
}
- // 64 / (c->dstBpc & ~7) is the same as 16 / sizeof(scaling_intermediate)
- c->uv_off_px = dst_stride_px + 64 / (c->dstBpc &~ 7);
- c->uv_off_byte = dst_stride + 16;
+ // 64 / c->scalingBpp is the same as 16 / sizeof(scaling_intermediate)
+ c->uv_off = (dst_stride>>1) + 64 / (c->dstBpc &~ 7);
+ c->uv_offx2 = dst_stride + 16;
for (i=0; i<c->vChrBufSize; i++) {
FF_ALLOC_OR_GOTO(c, c->chrUPixBuf[i+c->vChrBufSize], dst_stride*2+32, fail);
c->chrUPixBuf[i] = c->chrUPixBuf[i+c->vChrBufSize];
//try to avoid drawing green stuff between the right end and the stride end
for (i=0; i<c->vChrBufSize; i++)
- memset(c->chrUPixBuf[i], 64, dst_stride*2+1);
+ if(av_pix_fmt_descriptors[c->dstFormat].comp[0].depth_minus1 == 15){
+ av_assert0(c->dstBpc > 10);
+ for(j=0; j<dst_stride/2+1; j++)
+ ((int32_t*)(c->chrUPixBuf[i]))[j] = 1<<18;
+ } else
+ for(j=0; j<dst_stride+1; j++)
+ ((int16_t*)(c->chrUPixBuf[i]))[j] = 1<<14;
assert(c->chrDstH <= dstH);
else av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");
av_log(c, AV_LOG_INFO, "from %s to %s%s ",
- sws_format_name(srcFormat),
+ av_get_pix_fmt_name(srcFormat),
#ifdef DITHER1XBPP
dstFormat == PIX_FMT_BGR555 || dstFormat == PIX_FMT_BGR565 ||
dstFormat == PIX_FMT_RGB444BE || dstFormat == PIX_FMT_RGB444LE ||
#else
"",
#endif
- sws_format_name(dstFormat));
+ av_get_pix_fmt_name(dstFormat));
if (HAVE_MMX2 && cpu_flags & AV_CPU_FLAG_MMX2) av_log(c, AV_LOG_INFO, "using MMX2\n");
else if (HAVE_AMD3DNOW && cpu_flags & AV_CPU_FLAG_3DNOW) av_log(c, AV_LOG_INFO, "using 3DNOW\n");
#endif /* HAVE_MMX */
av_freep(&c->yuvTable);
- av_free(c->formatConvBuffer);
+ av_freep(&c->formatConvBuffer);
av_free(c);
}