From 0720d263ea2bdd28cd3be1d1e802d49d7c680b8c Mon Sep 17 00:00:00 2001 From: Justin Ruggles Date: Sat, 10 Mar 2012 18:50:45 -0500 Subject: [PATCH 1/7] FATE: use $fuzz directly in do_tiny_psnr() instead of passing it around --- tests/fate-run.sh | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/tests/fate-run.sh b/tests/fate-run.sh index a1503aecbd..419a3ebc8a 100755 --- a/tests/fate-run.sh +++ b/tests/fate-run.sh @@ -14,7 +14,7 @@ target_path=$4 command=$5 cmp=${6:-diff} ref=${7:-"${base}/ref/fate/${test}"} -fuzz=$8 +fuzz=${8:-1} threads=${9:-1} thread_type=${10:-frame+slice} cpuflags=${11:-all} @@ -30,7 +30,7 @@ do_tiny_psnr(){ val=$(expr "$psnr" : ".*$3: *\([0-9.]*\)") size1=$(expr "$psnr" : '.*bytes: *\([0-9]*\)') size2=$(expr "$psnr" : '.*bytes:[ 0-9]*/ *\([0-9]*\)') - res=$(echo "if ($val $4 $5) 1" | bc) + res=$(echo "if ($val $4 $fuzz) 1" | bc) if [ "$res" != 1 ] || [ $size1 != $size2 ]; then echo "$psnr" return 1 @@ -38,11 +38,11 @@ do_tiny_psnr(){ } oneoff(){ - do_tiny_psnr "$1" "$2" MAXDIFF '<=' ${fuzz:-1} + do_tiny_psnr "$1" "$2" MAXDIFF '<=' } stddev(){ - do_tiny_psnr "$1" "$2" stddev '<=' ${fuzz:-1} + do_tiny_psnr "$1" "$2" stddev '<=' } run(){ @@ -122,8 +122,8 @@ fi if test -e "$ref"; then case $cmp in diff) diff -u -w "$ref" "$outfile" >$cmpfile ;; - oneoff) oneoff "$ref" "$outfile" "$fuzz" >$cmpfile ;; - stddev) stddev "$ref" "$outfile" "$fuzz" >$cmpfile ;; + oneoff) oneoff "$ref" "$outfile" >$cmpfile ;; + stddev) stddev "$ref" "$outfile" >$cmpfile ;; null) cat "$outfile" >$cmpfile ;; esac cmperr=$? From 5ecadc662042f3d005a1a01618f448c9b1c44afc Mon Sep 17 00:00:00 2001 From: Justin Ruggles Date: Sat, 10 Mar 2012 18:53:19 -0500 Subject: [PATCH 2/7] FATE: allow tests to set CMP_SHIFT to pass to tiny_psnr This will allow adjusting for any encoder or decoder delay when doing comparisons. --- tests/Makefile | 2 +- tests/fate-run.sh | 3 ++- 2 files changed, 3 insertions(+), 2 deletions(-) diff --git a/tests/Makefile b/tests/Makefile index 5c65237b44..760915afe9 100644 --- a/tests/Makefile +++ b/tests/Makefile @@ -116,7 +116,7 @@ fate: $(FATE) $(FATE): avconv$(EXESUF) $(FATE_UTILS:%=tests/%$(HOSTEXESUF)) @echo "TEST $(@:fate-%=%)" - $(Q)$(SRC_PATH)/tests/fate-run.sh $@ "$(SAMPLES)" "$(TARGET_EXEC)" "$(TARGET_PATH)" '$(CMD)' '$(CMP)' '$(REF)' '$(FUZZ)' '$(THREADS)' '$(THREAD_TYPE)' '$(CPUFLAGS)' + $(Q)$(SRC_PATH)/tests/fate-run.sh $@ "$(SAMPLES)" "$(TARGET_EXEC)" "$(TARGET_PATH)" '$(CMD)' '$(CMP)' '$(REF)' '$(FUZZ)' '$(THREADS)' '$(THREAD_TYPE)' '$(CPUFLAGS)' '$(CMP_SHIFT)' fate-list: @printf '%s\n' $(sort $(FATE)) diff --git a/tests/fate-run.sh b/tests/fate-run.sh index 419a3ebc8a..94d7cf8a0c 100755 --- a/tests/fate-run.sh +++ b/tests/fate-run.sh @@ -18,6 +18,7 @@ fuzz=${8:-1} threads=${9:-1} thread_type=${10:-frame+slice} cpuflags=${11:-all} +cmp_shift=${12:-0} outdir="tests/data/fate" outfile="${outdir}/${test}" @@ -26,7 +27,7 @@ cmpfile="${outdir}/${test}.diff" repfile="${outdir}/${test}.rep" do_tiny_psnr(){ - psnr=$(tests/tiny_psnr "$1" "$2" 2 0 0) + psnr=$(tests/tiny_psnr "$1" "$2" 2 $cmp_shift 0) val=$(expr "$psnr" : ".*$3: *\([0-9.]*\)") size1=$(expr "$psnr" : '.*bytes: *\([0-9]*\)') size2=$(expr "$psnr" : '.*bytes:[ 0-9]*/ *\([0-9]*\)') From bb6842966e8b2c942a1853ade3d9bc7817dfc2ee Mon Sep 17 00:00:00 2001 From: Justin Ruggles Date: Sat, 10 Mar 2012 19:33:34 -0500 Subject: [PATCH 3/7] FATE: use absolute difference from a target value in do_tiny_psnr() This will allow comparison to original pre-encoded content instead of comparing to expected decoded output. --- tests/Makefile | 2 +- tests/fate-run.sh | 16 ++++++++++++---- 2 files changed, 13 insertions(+), 5 deletions(-) diff --git a/tests/Makefile b/tests/Makefile index 760915afe9..82eb67c447 100644 --- a/tests/Makefile +++ b/tests/Makefile @@ -116,7 +116,7 @@ fate: $(FATE) $(FATE): avconv$(EXESUF) $(FATE_UTILS:%=tests/%$(HOSTEXESUF)) @echo "TEST $(@:fate-%=%)" - $(Q)$(SRC_PATH)/tests/fate-run.sh $@ "$(SAMPLES)" "$(TARGET_EXEC)" "$(TARGET_PATH)" '$(CMD)' '$(CMP)' '$(REF)' '$(FUZZ)' '$(THREADS)' '$(THREAD_TYPE)' '$(CPUFLAGS)' '$(CMP_SHIFT)' + $(Q)$(SRC_PATH)/tests/fate-run.sh $@ "$(SAMPLES)" "$(TARGET_EXEC)" "$(TARGET_PATH)" '$(CMD)' '$(CMP)' '$(REF)' '$(FUZZ)' '$(THREADS)' '$(THREAD_TYPE)' '$(CPUFLAGS)' '$(CMP_SHIFT)' '$(CMP_TARGET)' fate-list: @printf '%s\n' $(sort $(FATE)) diff --git a/tests/fate-run.sh b/tests/fate-run.sh index 94d7cf8a0c..07065c3f30 100755 --- a/tests/fate-run.sh +++ b/tests/fate-run.sh @@ -19,6 +19,7 @@ threads=${9:-1} thread_type=${10:-frame+slice} cpuflags=${11:-all} cmp_shift=${12:-0} +cmp_target=${13:-0} outdir="tests/data/fate" outfile="${outdir}/${test}" @@ -26,24 +27,31 @@ errfile="${outdir}/${test}.err" cmpfile="${outdir}/${test}.diff" repfile="${outdir}/${test}.rep" +# $1=value1, $2=value2, $3=threshold +# prints 0 if absolute difference between value1 and value2 is <= threshold +compare(){ + v=$(echo "scale=2; if ($1 >= $2) { $1 - $2 } else { $2 - $1 }" | bc) + echo "if ($v <= $3) { 0 } else { 1 }" | bc +} + do_tiny_psnr(){ psnr=$(tests/tiny_psnr "$1" "$2" 2 $cmp_shift 0) val=$(expr "$psnr" : ".*$3: *\([0-9.]*\)") size1=$(expr "$psnr" : '.*bytes: *\([0-9]*\)') size2=$(expr "$psnr" : '.*bytes:[ 0-9]*/ *\([0-9]*\)') - res=$(echo "if ($val $4 $fuzz) 1" | bc) - if [ "$res" != 1 ] || [ $size1 != $size2 ]; then + val_cmp=$(compare $val $cmp_target $fuzz) + if [ "$val_cmp" != 0 ] || [ $size1 != $size2 ]; then echo "$psnr" return 1 fi } oneoff(){ - do_tiny_psnr "$1" "$2" MAXDIFF '<=' + do_tiny_psnr "$1" "$2" MAXDIFF } stddev(){ - do_tiny_psnr "$1" "$2" stddev '<=' + do_tiny_psnr "$1" "$2" stddev } run(){ From cffd7580bbdf29ae5b1e56af55ff21e8c85bba6c Mon Sep 17 00:00:00 2001 From: Justin Ruggles Date: Sat, 10 Mar 2012 19:52:05 -0500 Subject: [PATCH 4/7] FATE: allow a tolerance in the size comparison in do_tiny_psnr() This will allow for comparing decoded output to the original source when the decoded size is not exactly the same as the original size. --- tests/Makefile | 2 +- tests/fate-run.sh | 4 +++- 2 files changed, 4 insertions(+), 2 deletions(-) diff --git a/tests/Makefile b/tests/Makefile index 82eb67c447..4c21a6a5a4 100644 --- a/tests/Makefile +++ b/tests/Makefile @@ -116,7 +116,7 @@ fate: $(FATE) $(FATE): avconv$(EXESUF) $(FATE_UTILS:%=tests/%$(HOSTEXESUF)) @echo "TEST $(@:fate-%=%)" - $(Q)$(SRC_PATH)/tests/fate-run.sh $@ "$(SAMPLES)" "$(TARGET_EXEC)" "$(TARGET_PATH)" '$(CMD)' '$(CMP)' '$(REF)' '$(FUZZ)' '$(THREADS)' '$(THREAD_TYPE)' '$(CPUFLAGS)' '$(CMP_SHIFT)' '$(CMP_TARGET)' + $(Q)$(SRC_PATH)/tests/fate-run.sh $@ "$(SAMPLES)" "$(TARGET_EXEC)" "$(TARGET_PATH)" '$(CMD)' '$(CMP)' '$(REF)' '$(FUZZ)' '$(THREADS)' '$(THREAD_TYPE)' '$(CPUFLAGS)' '$(CMP_SHIFT)' '$(CMP_TARGET)' '$(SIZE_TOLERANCE)' fate-list: @printf '%s\n' $(sort $(FATE)) diff --git a/tests/fate-run.sh b/tests/fate-run.sh index 07065c3f30..40a4fe2f6d 100755 --- a/tests/fate-run.sh +++ b/tests/fate-run.sh @@ -20,6 +20,7 @@ thread_type=${10:-frame+slice} cpuflags=${11:-all} cmp_shift=${12:-0} cmp_target=${13:-0} +size_tolerance=${14:-0} outdir="tests/data/fate" outfile="${outdir}/${test}" @@ -40,7 +41,8 @@ do_tiny_psnr(){ size1=$(expr "$psnr" : '.*bytes: *\([0-9]*\)') size2=$(expr "$psnr" : '.*bytes:[ 0-9]*/ *\([0-9]*\)') val_cmp=$(compare $val $cmp_target $fuzz) - if [ "$val_cmp" != 0 ] || [ $size1 != $size2 ]; then + size_cmp=$(compare $size1 $size2 $size_tolerance) + if [ "$val_cmp" != 0 ] || [ "$size_cmp" != 0 ]; then echo "$psnr" return 1 fi From 90e5b58a53d863a43a86eea81295097dae13b25a Mon Sep 17 00:00:00 2001 From: Justin Ruggles Date: Sat, 10 Mar 2012 16:37:41 -0500 Subject: [PATCH 5/7] FATE: add capability for audio encode/decode tests with fuzzy psnr comparison This allows for testing floating-point audio encoders across different platforms where exact comparisons are unreliable due to float rounding differences. --- tests/fate-run.sh | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/tests/fate-run.sh b/tests/fate-run.sh index 40a4fe2f6d..dcfbc842da 100755 --- a/tests/fate-run.sh +++ b/tests/fate-run.sh @@ -85,6 +85,16 @@ pcm(){ avconv "$@" -vn -f s16le - } +enc_dec_pcm(){ + out_fmt=$1 + pcm_fmt=$2 + shift 2 + encfile="${outdir}/${test}.${out_fmt}" + cleanfiles=$encfile + avconv -i $ref "$@" -f $out_fmt -y $encfile || return + avconv -i $encfile -c:a pcm_${pcm_fmt} -f wav - +} + regtest(){ t="${test#$2-}" ref=${base}/ref/$2/$t From ad4cec6b80d2dcca5c95f9323abc706942bece63 Mon Sep 17 00:00:00 2001 From: Diego Biurrun Date: Fri, 16 Mar 2012 18:50:15 +0100 Subject: [PATCH 6/7] s3tc.h: Add missing #include to fix standalone header compilation. --- libavcodec/s3tc.h | 2 ++ 1 file changed, 2 insertions(+) diff --git a/libavcodec/s3tc.h b/libavcodec/s3tc.h index d48fb41eb3..25237b95bd 100644 --- a/libavcodec/s3tc.h +++ b/libavcodec/s3tc.h @@ -24,6 +24,8 @@ #include +#include "bytestream.h" + #define FF_S3TC_DXT1 0x31545844 #define FF_S3TC_DXT3 0x33545844 From e5d403720ec4914169f55913a5a5555d908500b6 Mon Sep 17 00:00:00 2001 From: Diego Biurrun Date: Fri, 16 Mar 2012 18:40:05 +0100 Subject: [PATCH 7/7] h264: K&R formatting cosmetics Also remove some disabled code and fix a few comment typos. --- libavcodec/h264.c | 4502 ++++++++++++++++++++++++--------------------- 1 file changed, 2422 insertions(+), 2080 deletions(-) diff --git a/libavcodec/h264.c b/libavcodec/h264.c index 64f676c4a9..0b7d6c1bb2 100644 --- a/libavcodec/h264.c +++ b/libavcodec/h264.c @@ -42,15 +42,19 @@ #include "vdpau_internal.h" #include "libavutil/avassert.h" -//#undef NDEBUG +// #undef NDEBUG #include -static const uint8_t rem6[QP_MAX_NUM+1]={ -0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, +static const uint8_t rem6[QP_MAX_NUM + 1] = { + 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, + 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, + 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, }; -static const uint8_t div6[QP_MAX_NUM+1]={ -0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 9, 9, 9, 9, 9, 9,10,10,10,10, +static const uint8_t div6[QP_MAX_NUM + 1] = { + 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 3, 3, 3, + 3, 3, 3, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, + 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 9, 9, 9, 9, 9, 9, 10, 10, 10, 10, }; static const enum PixelFormat hwaccel_pixfmt_list_h264_jpeg_420[] = { @@ -65,71 +69,89 @@ static const enum PixelFormat hwaccel_pixfmt_list_h264_jpeg_420[] = { * Check if the top & left blocks are available if needed and * change the dc mode so it only uses the available blocks. */ -int ff_h264_check_intra4x4_pred_mode(H264Context *h){ - MpegEncContext * const s = &h->s; - static const int8_t top [12]= {-1, 0,LEFT_DC_PRED,-1,-1,-1,-1,-1, 0}; - static const int8_t left[12]= { 0,-1, TOP_DC_PRED, 0,-1,-1,-1, 0,-1,DC_128_PRED}; +int ff_h264_check_intra4x4_pred_mode(H264Context *h) +{ + MpegEncContext *const s = &h->s; + static const int8_t top[12] = { + -1, 0, LEFT_DC_PRED, -1, -1, -1, -1, -1, 0 + }; + static const int8_t left[12] = { + 0, -1, TOP_DC_PRED, 0, -1, -1, -1, 0, -1, DC_128_PRED + }; int i; - if(!(h->top_samples_available&0x8000)){ - for(i=0; i<4; i++){ - int status= top[ h->intra4x4_pred_mode_cache[scan8[0] + i] ]; - if(status<0){ - av_log(h->s.avctx, AV_LOG_ERROR, "top block unavailable for requested intra4x4 mode %d at %d %d\n", status, s->mb_x, s->mb_y); + if (!(h->top_samples_available & 0x8000)) { + for (i = 0; i < 4; i++) { + int status = top[h->intra4x4_pred_mode_cache[scan8[0] + i]]; + if (status < 0) { + av_log(h->s.avctx, AV_LOG_ERROR, + "top block unavailable for requested intra4x4 mode %d at %d %d\n", + status, s->mb_x, s->mb_y); return -1; - } else if(status){ - h->intra4x4_pred_mode_cache[scan8[0] + i]= status; + } else if (status) { + h->intra4x4_pred_mode_cache[scan8[0] + i] = status; } } } - if((h->left_samples_available&0x8888)!=0x8888){ - static const int mask[4]={0x8000,0x2000,0x80,0x20}; - for(i=0; i<4; i++){ - if(!(h->left_samples_available&mask[i])){ - int status= left[ h->intra4x4_pred_mode_cache[scan8[0] + 8*i] ]; - if(status<0){ - av_log(h->s.avctx, AV_LOG_ERROR, "left block unavailable for requested intra4x4 mode %d at %d %d\n", status, s->mb_x, s->mb_y); + if ((h->left_samples_available & 0x8888) != 0x8888) { + static const int mask[4] = { 0x8000, 0x2000, 0x80, 0x20 }; + for (i = 0; i < 4; i++) + if (!(h->left_samples_available & mask[i])) { + int status = left[h->intra4x4_pred_mode_cache[scan8[0] + 8 * i]]; + if (status < 0) { + av_log(h->s.avctx, AV_LOG_ERROR, + "left block unavailable for requested intra4x4 mode %d at %d %d\n", + status, s->mb_x, s->mb_y); return -1; - } else if(status){ - h->intra4x4_pred_mode_cache[scan8[0] + 8*i]= status; + } else if (status) { + h->intra4x4_pred_mode_cache[scan8[0] + 8 * i] = status; } } - } } return 0; -} //FIXME cleanup like ff_h264_check_intra_pred_mode +} // FIXME cleanup like ff_h264_check_intra_pred_mode /** * Check if the top & left blocks are available if needed and * change the dc mode so it only uses the available blocks. */ -int ff_h264_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}; +int ff_h264_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(mode > 6U) { - av_log(h->s.avctx, AV_LOG_ERROR, "out of range intra chroma pred mode at %d %d\n", s->mb_x, s->mb_y); + if (mode > 6U) { + av_log(h->s.avctx, AV_LOG_ERROR, + "out of range intra chroma pred mode at %d %d\n", + s->mb_x, s->mb_y); return -1; } - if(!(h->top_samples_available&0x8000)){ - mode= top[ mode ]; - if(mode<0){ - av_log(h->s.avctx, AV_LOG_ERROR, "top block unavailable for requested intra mode at %d %d\n", s->mb_x, s->mb_y); + if (!(h->top_samples_available & 0x8000)) { + mode = top[mode]; + if (mode < 0) { + av_log(h->s.avctx, AV_LOG_ERROR, + "top block unavailable for requested intra mode at %d %d\n", + s->mb_x, s->mb_y); return -1; } } - if((h->left_samples_available&0x8080) != 0x8080){ - mode= left[ mode ]; - 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 ((h->left_samples_available & 0x8080) != 0x8080) { + mode = left[mode]; + 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){ - av_log(h->s.avctx, AV_LOG_ERROR, "left block unavailable for requested intra mode at %d %d\n", s->mb_x, s->mb_y); + if (mode < 0) { + av_log(h->s.avctx, AV_LOG_ERROR, + "left block unavailable for requested intra mode at %d %d\n", + s->mb_x, s->mb_y); return -1; } } @@ -137,89 +159,102 @@ int ff_h264_check_intra_pred_mode(H264Context *h, int mode, int is_chroma){ return mode; } -const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src, int *dst_length, int *consumed, int length){ +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; int bufidx; -// src[0]&0x80; //forbidden bit - h->nal_ref_idc= src[0]>>5; - h->nal_unit_type= src[0]&0x1F; + // src[0]&0x80; // forbidden bit + h->nal_ref_idc = src[0] >> 5; + h->nal_unit_type = src[0] & 0x1F; - src++; length--; + src++; + length--; #if HAVE_FAST_UNALIGNED -# if HAVE_FAST_64BIT -# define RS 7 - for(i=0; i+10 && !src[i]) i--; - while(src[i]) i++; +#if HAVE_FAST_64BIT +#define RS 7 + for (i = 0; i + 1 < length; i += 9) { + if (!((~AV_RN64A(src + i) & + (AV_RN64A(src + i) - 0x0100010001000101ULL)) & + 0x8000800080008080ULL)) #else -# define RS 0 - for(i=0; i+10 && src[i-1]==0) i--; +#define RS 3 + for (i = 0; i + 1 < length; i += 5) { + if (!((~AV_RN32A(src + i) & + (AV_RN32A(src + i) - 0x01000101U)) & + 0x80008080U)) #endif - if(i+2 0 && !src[i]) + i--; + while (src[i]) + i++; +#else +#define RS 0 + for (i = 0; i + 1 < length; i += 2) { + if (src[i]) + continue; + if (i > 0 && src[i - 1] == 0) + i--; +#endif + if (i + 2 < length && src[i + 1] == 0 && src[i + 2] <= 3) { + if (src[i + 2] != 3) { /* startcode, so we must be past the end */ - length=i; + length = i; } break; } - i-= RS; + i -= RS; } - if(i>=length-1){ //no escaped 0 - *dst_length= length; - *consumed= length+1; //+1 for the header + 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); - dst= h->rbsp_buffer[bufidx]; + // use second escape buffer for inter data + bufidx = h->nal_unit_type == NAL_DPC ? 1 : 0; + av_fast_malloc(&h->rbsp_buffer[bufidx], &h->rbsp_buffer_size[bufidx], + length + FF_INPUT_BUFFER_PADDING_SIZE); + dst = h->rbsp_buffer[bufidx]; - if (dst == NULL){ + if (dst == NULL) return NULL; - } -//printf("decoding esc\n"); + // printf("decoding esc\n"); memcpy(dst, src, i); - si=di=i; - while(si+23){ - dst[di++]= src[si++]; - dst[di++]= src[si++]; - }else if(src[si]==0 && src[si+1]==0){ - if(src[si+2]==3){ //escape - dst[di++]= 0; - dst[di++]= 0; - si+=3; + si = di = i; + while (si + 2 < length) { + // remove escapes (very rare 1:2^22) + if (src[si + 2] > 3) { + dst[di++] = src[si++]; + dst[di++] = src[si++]; + } else if (src[si] == 0 && src[si + 1] == 0) { + if (src[si + 2] == 3) { // escape + dst[di++] = 0; + dst[di++] = 0; + si += 3; continue; - }else //next start code + } else // next start code goto nsc; } - dst[di++]= src[si++]; + dst[di++] = src[si++]; } - while(sis.avctx, "rbsp trailing %X\n", v); - for(r=1; r<9; r++){ - if(v&1) return r; - v>>=1; + for (r = 1; r < 9; r++) { + if (v & 1) + return r; + v >>= 1; } return 0; } -static inline int get_lowest_part_list_y(H264Context *h, Picture *pic, int n, int height, - int y_offset, int list){ - int raw_my= h->mv_cache[list][ scan8[n] ][1]; - int filter_height= (raw_my&3) ? 2 : 0; - int full_my= (raw_my>>2) + y_offset; - int top = full_my - filter_height, bottom = full_my + height + filter_height; +static inline int get_lowest_part_list_y(H264Context *h, Picture *pic, int n, + int height, int y_offset, int list) +{ + int raw_my = h->mv_cache[list][scan8[n]][1]; + int filter_height = (raw_my & 3) ? 2 : 0; + int full_my = (raw_my >> 2) + y_offset; + int top = full_my - filter_height; + int bottom = full_my + filter_height + height; return FFMAX(abs(top), bottom); } -static inline void get_lowest_part_y(H264Context *h, int refs[2][48], int n, int height, - int y_offset, int list0, int list1, int *nrefs){ - MpegEncContext * const s = &h->s; +static inline void get_lowest_part_y(H264Context *h, int refs[2][48], int n, + int height, int y_offset, int list0, + int list1, int *nrefs) +{ + MpegEncContext *const s = &h->s; int my; - y_offset += 16*(s->mb_y >> MB_FIELD); + y_offset += 16 * (s->mb_y >> MB_FIELD); - if(list0){ - int ref_n = h->ref_cache[0][ scan8[n] ]; - Picture *ref= &h->ref_list[0][ref_n]; + if (list0) { + int ref_n = h->ref_cache[0][scan8[n]]; + Picture *ref = &h->ref_list[0][ref_n]; // Error resilience puts the current picture in the ref list. // Don't try to wait on these as it will cause a deadlock. // Fields can wait on each other, though. - if (ref->f.thread_opaque != s->current_picture.f.thread_opaque || - (ref->f.reference & 3) != s->picture_structure) { + if (ref->f.thread_opaque != s->current_picture.f.thread_opaque || + (ref->f.reference & 3) != s->picture_structure) { my = get_lowest_part_list_y(h, ref, n, height, y_offset, 0); - if (refs[0][ref_n] < 0) nrefs[0] += 1; + if (refs[0][ref_n] < 0) + nrefs[0] += 1; refs[0][ref_n] = FFMAX(refs[0][ref_n], my); } } - if(list1){ - int ref_n = h->ref_cache[1][ scan8[n] ]; - Picture *ref= &h->ref_list[1][ref_n]; + if (list1) { + int ref_n = h->ref_cache[1][scan8[n]]; + Picture *ref = &h->ref_list[1][ref_n]; - if (ref->f.thread_opaque != s->current_picture.f.thread_opaque || - (ref->f.reference & 3) != s->picture_structure) { + if (ref->f.thread_opaque != s->current_picture.f.thread_opaque || + (ref->f.reference & 3) != s->picture_structure) { my = get_lowest_part_list_y(h, ref, n, height, y_offset, 1); - if (refs[1][ref_n] < 0) nrefs[1] += 1; + if (refs[1][ref_n] < 0) + nrefs[1] += 1; refs[1][ref_n] = FFMAX(refs[1][ref_n], my); } } @@ -290,337 +333,321 @@ static inline void get_lowest_part_y(H264Context *h, int refs[2][48], int n, int * * @param h the H264 context */ -static void await_references(H264Context *h){ - MpegEncContext * const s = &h->s; - const int mb_xy= h->mb_xy; +static void await_references(H264Context *h) +{ + MpegEncContext *const s = &h->s; + const int mb_xy = h->mb_xy; const int mb_type = s->current_picture.f.mb_type[mb_xy]; int refs[2][48]; - int nrefs[2] = {0}; + int nrefs[2] = { 0 }; int ref, list; memset(refs, -1, sizeof(refs)); - if(IS_16X16(mb_type)){ + if (IS_16X16(mb_type)) { get_lowest_part_y(h, refs, 0, 16, 0, - IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1), nrefs); - }else if(IS_16X8(mb_type)){ + IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1), nrefs); + } else if (IS_16X8(mb_type)) { get_lowest_part_y(h, refs, 0, 8, 0, - IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1), nrefs); + IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1), nrefs); get_lowest_part_y(h, refs, 8, 8, 8, - IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1), nrefs); - }else if(IS_8X16(mb_type)){ + IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1), nrefs); + } else if (IS_8X16(mb_type)) { get_lowest_part_y(h, refs, 0, 16, 0, - IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1), nrefs); + IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1), nrefs); get_lowest_part_y(h, refs, 4, 16, 0, - IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1), nrefs); - }else{ + IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1), nrefs); + } else { int i; assert(IS_8X8(mb_type)); - for(i=0; i<4; i++){ - const int sub_mb_type= h->sub_mb_type[i]; - const int n= 4*i; - int y_offset= (i&2)<<2; + for (i = 0; i < 4; i++) { + const int sub_mb_type = h->sub_mb_type[i]; + const int n = 4 * i; + int y_offset = (i & 2) << 2; - if(IS_SUB_8X8(sub_mb_type)){ - get_lowest_part_y(h, refs, n , 8, y_offset, - IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), nrefs); - }else if(IS_SUB_8X4(sub_mb_type)){ - get_lowest_part_y(h, refs, n , 4, y_offset, - IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), nrefs); - get_lowest_part_y(h, refs, n+2, 4, y_offset+4, - IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), nrefs); - }else if(IS_SUB_4X8(sub_mb_type)){ - get_lowest_part_y(h, refs, n , 8, y_offset, - IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), nrefs); - get_lowest_part_y(h, refs, n+1, 8, y_offset, - IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), nrefs); - }else{ + if (IS_SUB_8X8(sub_mb_type)) { + get_lowest_part_y(h, refs, n, 8, y_offset, + IS_DIR(sub_mb_type, 0, 0), + IS_DIR(sub_mb_type, 0, 1), + nrefs); + } else if (IS_SUB_8X4(sub_mb_type)) { + get_lowest_part_y(h, refs, n, 4, y_offset, + IS_DIR(sub_mb_type, 0, 0), + IS_DIR(sub_mb_type, 0, 1), + nrefs); + get_lowest_part_y(h, refs, n + 2, 4, y_offset + 4, + IS_DIR(sub_mb_type, 0, 0), + IS_DIR(sub_mb_type, 0, 1), + nrefs); + } else if (IS_SUB_4X8(sub_mb_type)) { + get_lowest_part_y(h, refs, n, 8, y_offset, + IS_DIR(sub_mb_type, 0, 0), + IS_DIR(sub_mb_type, 0, 1), + nrefs); + get_lowest_part_y(h, refs, n + 1, 8, y_offset, + IS_DIR(sub_mb_type, 0, 0), + IS_DIR(sub_mb_type, 0, 1), + nrefs); + } else { int j; assert(IS_SUB_4X4(sub_mb_type)); - for(j=0; j<4; j++){ - int sub_y_offset= y_offset + 2*(j&2); - get_lowest_part_y(h, refs, n+j, 4, sub_y_offset, - IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), nrefs); + for (j = 0; j < 4; j++) { + int sub_y_offset = y_offset + 2 * (j & 2); + get_lowest_part_y(h, refs, n + j, 4, sub_y_offset, + IS_DIR(sub_mb_type, 0, 0), + IS_DIR(sub_mb_type, 0, 1), + nrefs); } } } } - for(list=h->list_count-1; list>=0; list--){ - for(ref=0; ref<48 && nrefs[list]; ref++){ + for (list = h->list_count - 1; list >= 0; list--) + for (ref = 0; ref < 48 && nrefs[list]; ref++) { int row = refs[list][ref]; - if(row >= 0){ - Picture *ref_pic = &h->ref_list[list][ref]; - int ref_field = ref_pic->f.reference - 1; + if (row >= 0) { + Picture *ref_pic = &h->ref_list[list][ref]; + int ref_field = ref_pic->f.reference - 1; int ref_field_picture = ref_pic->field_picture; - int pic_height = 16*s->mb_height >> ref_field_picture; + int pic_height = 16 * s->mb_height >> ref_field_picture; row <<= MB_MBAFF; nrefs[list]--; - if(!FIELD_PICTURE && ref_field_picture){ // frame referencing two fields - ff_thread_await_progress(&ref_pic->f, FFMIN((row >> 1) - !(row & 1), pic_height - 1), 1); - ff_thread_await_progress(&ref_pic->f, FFMIN((row >> 1), pic_height - 1), 0); - }else if(FIELD_PICTURE && !ref_field_picture){ // field referencing one field of a frame - ff_thread_await_progress(&ref_pic->f, FFMIN(row * 2 + ref_field, pic_height - 1), 0); - }else if(FIELD_PICTURE){ - ff_thread_await_progress(&ref_pic->f, FFMIN(row, pic_height - 1), ref_field); - }else{ - ff_thread_await_progress(&ref_pic->f, FFMIN(row, pic_height - 1), 0); + if (!FIELD_PICTURE && ref_field_picture) { // frame referencing two fields + ff_thread_await_progress(&ref_pic->f, + FFMIN((row >> 1) - !(row & 1), + pic_height - 1), + 1); + ff_thread_await_progress(&ref_pic->f, + FFMIN((row >> 1), pic_height - 1), + 0); + } else if (FIELD_PICTURE && !ref_field_picture) { // field referencing one field of a frame + ff_thread_await_progress(&ref_pic->f, + FFMIN(row * 2 + ref_field, + pic_height - 1), + 0); + } else if (FIELD_PICTURE) { + ff_thread_await_progress(&ref_pic->f, + FFMIN(row, pic_height - 1), + ref_field); + } else { + ff_thread_await_progress(&ref_pic->f, + FFMIN(row, pic_height - 1), + 0); } } } - } } -#if 0 -/** - * DCT transforms the 16 dc values. - * @param qp quantization parameter ??? FIXME - */ -static void h264_luma_dc_dct_c(DCTELEM *block/*, int qp*/){ -// const int qmul= dequant_coeff[qp][0]; - int i; - int temp[16]; //FIXME check if this is a good idea - static const int x_offset[4]={0, 1*stride, 4* stride, 5*stride}; - static const int y_offset[4]={0, 2*stride, 8* stride, 10*stride}; - - for(i=0; i<4; i++){ - const int offset= y_offset[i]; - const int z0= block[offset+stride*0] + block[offset+stride*4]; - const int z1= block[offset+stride*0] - block[offset+stride*4]; - const int z2= block[offset+stride*1] - block[offset+stride*5]; - const int z3= block[offset+stride*1] + block[offset+stride*5]; - - temp[4*i+0]= z0+z3; - temp[4*i+1]= z1+z2; - temp[4*i+2]= z1-z2; - temp[4*i+3]= z0-z3; - } - - for(i=0; i<4; i++){ - const int offset= x_offset[i]; - const int z0= temp[4*0+i] + temp[4*2+i]; - const int z1= temp[4*0+i] - temp[4*2+i]; - const int z2= temp[4*1+i] - temp[4*3+i]; - const int z3= temp[4*1+i] + temp[4*3+i]; - - block[stride*0 +offset]= (z0 + z3)>>1; - block[stride*2 +offset]= (z1 + z2)>>1; - block[stride*8 +offset]= (z1 - z2)>>1; - block[stride*10+offset]= (z0 - z3)>>1; - } -} -#endif - -#undef xStride -#undef stride - -#if 0 -static void chroma_dc_dct_c(DCTELEM *block){ - const int stride= 16*2; - const int xStride= 16; - int a,b,c,d,e; - - a= block[stride*0 + xStride*0]; - b= block[stride*0 + xStride*1]; - c= block[stride*1 + xStride*0]; - d= block[stride*1 + xStride*1]; - - e= a-b; - a= a+b; - b= c-d; - c= c+d; - - block[stride*0 + xStride*0]= (a+c); - block[stride*0 + xStride*1]= (e+b); - block[stride*1 + xStride*0]= (a-c); - block[stride*1 + xStride*1]= (e-b); -} -#endif - -static av_always_inline void -mc_dir_part(H264Context *h, Picture *pic, int n, int square, - int height, int delta, int list, - uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, - int src_x_offset, int src_y_offset, - qpel_mc_func *qpix_op, h264_chroma_mc_func chroma_op, - int pixel_shift, int chroma_idc) +static av_always_inline void mc_dir_part(H264Context *h, Picture *pic, + int n, int square, int height, + int delta, int list, + uint8_t *dest_y, uint8_t *dest_cb, + uint8_t *dest_cr, + int src_x_offset, int src_y_offset, + qpel_mc_func *qpix_op, + h264_chroma_mc_func chroma_op, + int pixel_shift, int chroma_idc) { - MpegEncContext * const s = &h->s; - const int mx= h->mv_cache[list][ scan8[n] ][0] + src_x_offset*8; - int my= h->mv_cache[list][ scan8[n] ][1] + src_y_offset*8; - const int luma_xy= (mx&3) + ((my&3)<<2); - int offset = ((mx>>2) << pixel_shift) + (my>>2)*h->mb_linesize; - uint8_t * src_y = pic->f.data[0] + offset; - uint8_t * src_cb, * src_cr; - int extra_width= h->emu_edge_width; - int extra_height= h->emu_edge_height; - int emu=0; - const int full_mx= mx>>2; - const int full_my= my>>2; - const int pic_width = 16*s->mb_width; - const int pic_height = 16*s->mb_height >> MB_FIELD; + MpegEncContext *const s = &h->s; + const int mx = h->mv_cache[list][scan8[n]][0] + src_x_offset * 8; + int my = h->mv_cache[list][scan8[n]][1] + src_y_offset * 8; + const int luma_xy = (mx & 3) + ((my & 3) << 2); + int offset = ((mx >> 2) << pixel_shift) + (my >> 2) * h->mb_linesize; + uint8_t *src_y = pic->f.data[0] + offset; + uint8_t *src_cb, *src_cr; + int extra_width = h->emu_edge_width; + int extra_height = h->emu_edge_height; + int emu = 0; + const int full_mx = mx >> 2; + const int full_my = my >> 2; + const int pic_width = 16 * s->mb_width; + const int pic_height = 16 * s->mb_height >> MB_FIELD; int ysh; - if(mx&7) extra_width -= 3; - if(my&7) extra_height -= 3; + if (mx & 7) + extra_width -= 3; + if (my & 7) + extra_height -= 3; - if( full_mx < 0-extra_width - || full_my < 0-extra_height - || full_mx + 16/*FIXME*/ > pic_width + extra_width - || full_my + 16/*FIXME*/ > pic_height + extra_height){ - s->dsp.emulated_edge_mc(s->edge_emu_buffer, src_y - (2 << pixel_shift) - 2*h->mb_linesize, h->mb_linesize, - 16+5, 16+5/*FIXME*/, full_mx-2, full_my-2, pic_width, pic_height); - src_y= s->edge_emu_buffer + (2 << pixel_shift) + 2*h->mb_linesize; - emu=1; + if (full_mx < 0 - extra_width || + full_my < 0 - extra_height || + full_mx + 16 /*FIXME*/ > pic_width + extra_width || + full_my + 16 /*FIXME*/ > pic_height + extra_height) { + s->dsp.emulated_edge_mc(s->edge_emu_buffer, + src_y - (2 << pixel_shift) - 2 * h->mb_linesize, + h->mb_linesize, + 16 + 5, 16 + 5 /*FIXME*/, full_mx - 2, + full_my - 2, pic_width, pic_height); + src_y = s->edge_emu_buffer + (2 << pixel_shift) + 2 * h->mb_linesize; + emu = 1; } - qpix_op[luma_xy](dest_y, src_y, h->mb_linesize); //FIXME try variable height perhaps? - if(!square){ + qpix_op[luma_xy](dest_y, src_y, h->mb_linesize); // FIXME try variable height perhaps? + if (!square) qpix_op[luma_xy](dest_y + delta, src_y + delta, h->mb_linesize); - } - if(CONFIG_GRAY && s->flags&CODEC_FLAG_GRAY) return; + if (CONFIG_GRAY && s->flags & CODEC_FLAG_GRAY) + return; - if(chroma_idc == 3 /* yuv444 */){ + if (chroma_idc == 3 /* yuv444 */) { src_cb = pic->f.data[1] + offset; - if(emu){ - s->dsp.emulated_edge_mc(s->edge_emu_buffer, src_cb - (2 << pixel_shift) - 2*h->mb_linesize, h->mb_linesize, - 16+5, 16+5/*FIXME*/, full_mx-2, full_my-2, pic_width, pic_height); - src_cb= s->edge_emu_buffer + (2 << pixel_shift) + 2*h->mb_linesize; + if (emu) { + s->dsp.emulated_edge_mc(s->edge_emu_buffer, + src_cb - (2 << pixel_shift) - 2 * h->mb_linesize, + h->mb_linesize, + 16 + 5, 16 + 5 /*FIXME*/, + full_mx - 2, full_my - 2, + pic_width, pic_height); + src_cb = s->edge_emu_buffer + (2 << pixel_shift) + 2 * h->mb_linesize; } - qpix_op[luma_xy](dest_cb, src_cb, h->mb_linesize); //FIXME try variable height perhaps? - if(!square){ + qpix_op[luma_xy](dest_cb, src_cb, h->mb_linesize); // FIXME try variable height perhaps? + if (!square) qpix_op[luma_xy](dest_cb + delta, src_cb + delta, h->mb_linesize); - } src_cr = pic->f.data[2] + offset; - if(emu){ - s->dsp.emulated_edge_mc(s->edge_emu_buffer, src_cr - (2 << pixel_shift) - 2*h->mb_linesize, h->mb_linesize, - 16+5, 16+5/*FIXME*/, full_mx-2, full_my-2, pic_width, pic_height); - src_cr= s->edge_emu_buffer + (2 << pixel_shift) + 2*h->mb_linesize; + if (emu) { + s->dsp.emulated_edge_mc(s->edge_emu_buffer, + src_cr - (2 << pixel_shift) - 2 * h->mb_linesize, + h->mb_linesize, + 16 + 5, 16 + 5 /*FIXME*/, + full_mx - 2, full_my - 2, + pic_width, pic_height); + src_cr = s->edge_emu_buffer + (2 << pixel_shift) + 2 * h->mb_linesize; } - qpix_op[luma_xy](dest_cr, src_cr, h->mb_linesize); //FIXME try variable height perhaps? - if(!square){ + qpix_op[luma_xy](dest_cr, src_cr, h->mb_linesize); // FIXME try variable height perhaps? + if (!square) qpix_op[luma_xy](dest_cr + delta, src_cr + delta, h->mb_linesize); - } return; } ysh = 3 - (chroma_idc == 2 /* yuv422 */); - if(chroma_idc == 1 /* yuv420 */ && MB_FIELD){ + if (chroma_idc == 1 /* yuv420 */ && MB_FIELD) { // chroma offset when predicting from a field of opposite parity - my += 2 * ((s->mb_y & 1) - (pic->f.reference - 1)); - emu |= (my>>3) < 0 || (my>>3) + 8 >= (pic_height>>1); + my += 2 * ((s->mb_y & 1) - (pic->f.reference - 1)); + emu |= (my >> 3) < 0 || (my >> 3) + 8 >= (pic_height >> 1); } - src_cb = pic->f.data[1] + ((mx >> 3) << pixel_shift) + (my >> ysh) * h->mb_uvlinesize; - src_cr = pic->f.data[2] + ((mx >> 3) << pixel_shift) + (my >> ysh) * h->mb_uvlinesize; + src_cb = pic->f.data[1] + ((mx >> 3) << pixel_shift) + + (my >> ysh) * h->mb_uvlinesize; + src_cr = pic->f.data[2] + ((mx >> 3) << pixel_shift) + + (my >> ysh) * h->mb_uvlinesize; - if(emu){ + if (emu) { s->dsp.emulated_edge_mc(s->edge_emu_buffer, src_cb, h->mb_uvlinesize, 9, 8 * chroma_idc + 1, (mx >> 3), (my >> ysh), pic_width >> 1, pic_height >> (chroma_idc == 1 /* yuv420 */)); - src_cb= s->edge_emu_buffer; + src_cb = s->edge_emu_buffer; } - chroma_op(dest_cb, src_cb, h->mb_uvlinesize, height >> (chroma_idc == 1 /* yuv420 */), - mx&7, (my << (chroma_idc == 2 /* yuv422 */)) &7); + chroma_op(dest_cb, src_cb, h->mb_uvlinesize, + height >> (chroma_idc == 1 /* yuv420 */), + mx & 7, (my << (chroma_idc == 2 /* yuv422 */)) & 7); - if(emu){ + if (emu) { s->dsp.emulated_edge_mc(s->edge_emu_buffer, src_cr, h->mb_uvlinesize, 9, 8 * chroma_idc + 1, (mx >> 3), (my >> ysh), pic_width >> 1, pic_height >> (chroma_idc == 1 /* yuv420 */)); - src_cr= s->edge_emu_buffer; + src_cr = s->edge_emu_buffer; } chroma_op(dest_cr, src_cr, h->mb_uvlinesize, height >> (chroma_idc == 1 /* yuv420 */), - mx&7, (my << (chroma_idc == 2 /* yuv422 */)) &7); + mx & 7, (my << (chroma_idc == 2 /* yuv422 */)) & 7); } -static av_always_inline void -mc_part_std(H264Context *h, int n, int square, int height, int delta, - uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, - int x_offset, int y_offset, - qpel_mc_func *qpix_put, h264_chroma_mc_func chroma_put, - qpel_mc_func *qpix_avg, h264_chroma_mc_func chroma_avg, - int list0, int list1, int pixel_shift, int chroma_idc) +static av_always_inline void mc_part_std(H264Context *h, int n, int square, + int height, int delta, + uint8_t *dest_y, uint8_t *dest_cb, + uint8_t *dest_cr, + int x_offset, int y_offset, + qpel_mc_func *qpix_put, + h264_chroma_mc_func chroma_put, + qpel_mc_func *qpix_avg, + h264_chroma_mc_func chroma_avg, + int list0, int list1, + int pixel_shift, int chroma_idc) { - MpegEncContext * const s = &h->s; - qpel_mc_func *qpix_op= qpix_put; - h264_chroma_mc_func chroma_op= chroma_put; + MpegEncContext *const s = &h->s; + qpel_mc_func *qpix_op = qpix_put; + h264_chroma_mc_func chroma_op = chroma_put; - dest_y += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize; + dest_y += (2 * x_offset << pixel_shift) + 2 * y_offset * h->mb_linesize; if (chroma_idc == 3 /* yuv444 */) { - dest_cb += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize; - dest_cr += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize; + dest_cb += (2 * x_offset << pixel_shift) + 2 * y_offset * h->mb_linesize; + dest_cr += (2 * x_offset << pixel_shift) + 2 * y_offset * h->mb_linesize; } else if (chroma_idc == 2 /* yuv422 */) { - dest_cb += ( x_offset << pixel_shift) + 2*y_offset*h->mb_uvlinesize; - dest_cr += ( x_offset << pixel_shift) + 2*y_offset*h->mb_uvlinesize; - } else /* yuv420 */ { - dest_cb += ( x_offset << pixel_shift) + y_offset*h->mb_uvlinesize; - dest_cr += ( x_offset << pixel_shift) + y_offset*h->mb_uvlinesize; + dest_cb += (x_offset << pixel_shift) + 2 * y_offset * h->mb_uvlinesize; + dest_cr += (x_offset << pixel_shift) + 2 * y_offset * h->mb_uvlinesize; + } else { /* yuv420 */ + dest_cb += (x_offset << pixel_shift) + y_offset * h->mb_uvlinesize; + dest_cr += (x_offset << pixel_shift) + y_offset * h->mb_uvlinesize; } - x_offset += 8*s->mb_x; - y_offset += 8*(s->mb_y >> MB_FIELD); + x_offset += 8 * s->mb_x; + y_offset += 8 * (s->mb_y >> MB_FIELD); - if(list0){ - Picture *ref= &h->ref_list[0][ h->ref_cache[0][ scan8[n] ] ]; + if (list0) { + Picture *ref = &h->ref_list[0][h->ref_cache[0][scan8[n]]]; mc_dir_part(h, ref, n, square, height, delta, 0, - dest_y, dest_cb, dest_cr, x_offset, y_offset, - qpix_op, chroma_op, pixel_shift, chroma_idc); + dest_y, dest_cb, dest_cr, x_offset, y_offset, + qpix_op, chroma_op, pixel_shift, chroma_idc); - qpix_op= qpix_avg; - chroma_op= chroma_avg; + qpix_op = qpix_avg; + chroma_op = chroma_avg; } - if(list1){ - Picture *ref= &h->ref_list[1][ h->ref_cache[1][ scan8[n] ] ]; + if (list1) { + Picture *ref = &h->ref_list[1][h->ref_cache[1][scan8[n]]]; mc_dir_part(h, ref, n, square, height, delta, 1, - dest_y, dest_cb, dest_cr, x_offset, y_offset, - qpix_op, chroma_op, pixel_shift, chroma_idc); + dest_y, dest_cb, dest_cr, x_offset, y_offset, + qpix_op, chroma_op, pixel_shift, chroma_idc); } } -static av_always_inline void -mc_part_weighted(H264Context *h, int n, int square, int height, int delta, - uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, - int x_offset, int y_offset, - qpel_mc_func *qpix_put, h264_chroma_mc_func chroma_put, - h264_weight_func luma_weight_op, h264_weight_func chroma_weight_op, - h264_biweight_func luma_weight_avg, h264_biweight_func chroma_weight_avg, - int list0, int list1, int pixel_shift, int chroma_idc){ - MpegEncContext * const s = &h->s; +static av_always_inline void mc_part_weighted(H264Context *h, int n, int square, + int height, int delta, + uint8_t *dest_y, uint8_t *dest_cb, + uint8_t *dest_cr, + int x_offset, int y_offset, + qpel_mc_func *qpix_put, + h264_chroma_mc_func chroma_put, + h264_weight_func luma_weight_op, + h264_weight_func chroma_weight_op, + h264_biweight_func luma_weight_avg, + h264_biweight_func chroma_weight_avg, + int list0, int list1, + int pixel_shift, int chroma_idc) +{ + MpegEncContext *const s = &h->s; int chroma_height; - dest_y += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize; + dest_y += (2 * x_offset << pixel_shift) + 2 * y_offset * h->mb_linesize; if (chroma_idc == 3 /* yuv444 */) { - chroma_height = height; + chroma_height = height; chroma_weight_avg = luma_weight_avg; - chroma_weight_op = luma_weight_op; - dest_cb += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize; - dest_cr += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize; + chroma_weight_op = luma_weight_op; + dest_cb += (2 * x_offset << pixel_shift) + 2 * y_offset * h->mb_linesize; + dest_cr += (2 * x_offset << pixel_shift) + 2 * y_offset * h->mb_linesize; } else if (chroma_idc == 2 /* yuv422 */) { chroma_height = height; - dest_cb += ( x_offset << pixel_shift) + 2*y_offset*h->mb_uvlinesize; - dest_cr += ( x_offset << pixel_shift) + 2*y_offset*h->mb_uvlinesize; - } else /* yuv420 */ { + dest_cb += (x_offset << pixel_shift) + 2 * y_offset * h->mb_uvlinesize; + dest_cr += (x_offset << pixel_shift) + 2 * y_offset * h->mb_uvlinesize; + } else { /* yuv420 */ chroma_height = height >> 1; - dest_cb += ( x_offset << pixel_shift) + y_offset*h->mb_uvlinesize; - dest_cr += ( x_offset << pixel_shift) + y_offset*h->mb_uvlinesize; + dest_cb += (x_offset << pixel_shift) + y_offset * h->mb_uvlinesize; + dest_cr += (x_offset << pixel_shift) + y_offset * h->mb_uvlinesize; } - x_offset += 8*s->mb_x; - y_offset += 8*(s->mb_y >> MB_FIELD); + x_offset += 8 * s->mb_x; + y_offset += 8 * (s->mb_y >> MB_FIELD); - if(list0 && list1){ + if (list0 && list1) { /* don't optimize for luma-only case, since B-frames usually * use implicit weights => chroma too. */ uint8_t *tmp_cb = s->obmc_scratchpad; uint8_t *tmp_cr = s->obmc_scratchpad + (16 << pixel_shift); - uint8_t *tmp_y = s->obmc_scratchpad + 16*h->mb_uvlinesize; - int refn0 = h->ref_cache[0][ scan8[n] ]; - int refn1 = h->ref_cache[1][ scan8[n] ]; + uint8_t *tmp_y = s->obmc_scratchpad + 16 * h->mb_uvlinesize; + int refn0 = h->ref_cache[0][scan8[n]]; + int refn1 = h->ref_cache[1][scan8[n]]; mc_dir_part(h, &h->ref_list[0][refn0], n, square, height, delta, 0, dest_y, dest_cb, dest_cr, @@ -631,57 +658,77 @@ mc_part_weighted(H264Context *h, int n, int square, int height, int delta, x_offset, y_offset, qpix_put, chroma_put, pixel_shift, chroma_idc); - if(h->use_weight == 2){ - int weight0 = h->implicit_weight[refn0][refn1][s->mb_y&1]; + if (h->use_weight == 2) { + int weight0 = h->implicit_weight[refn0][refn1][s->mb_y & 1]; int weight1 = 64 - weight0; - luma_weight_avg( dest_y, tmp_y, h-> mb_linesize, - height, 5, weight0, weight1, 0); + luma_weight_avg(dest_y, tmp_y, h->mb_linesize, + height, 5, weight0, weight1, 0); chroma_weight_avg(dest_cb, tmp_cb, h->mb_uvlinesize, chroma_height, 5, weight0, weight1, 0); chroma_weight_avg(dest_cr, tmp_cr, h->mb_uvlinesize, chroma_height, 5, weight0, weight1, 0); - }else{ - luma_weight_avg(dest_y, tmp_y, h->mb_linesize, height, h->luma_log2_weight_denom, - h->luma_weight[refn0][0][0] , h->luma_weight[refn1][1][0], - h->luma_weight[refn0][0][1] + h->luma_weight[refn1][1][1]); - chroma_weight_avg(dest_cb, tmp_cb, h->mb_uvlinesize, chroma_height, h->chroma_log2_weight_denom, - h->chroma_weight[refn0][0][0][0] , h->chroma_weight[refn1][1][0][0], - h->chroma_weight[refn0][0][0][1] + h->chroma_weight[refn1][1][0][1]); - chroma_weight_avg(dest_cr, tmp_cr, h->mb_uvlinesize, chroma_height, h->chroma_log2_weight_denom, - h->chroma_weight[refn0][0][1][0] , h->chroma_weight[refn1][1][1][0], - h->chroma_weight[refn0][0][1][1] + h->chroma_weight[refn1][1][1][1]); + } else { + luma_weight_avg(dest_y, tmp_y, h->mb_linesize, height, + h->luma_log2_weight_denom, + h->luma_weight[refn0][0][0], + h->luma_weight[refn1][1][0], + h->luma_weight[refn0][0][1] + + h->luma_weight[refn1][1][1]); + chroma_weight_avg(dest_cb, tmp_cb, h->mb_uvlinesize, chroma_height, + h->chroma_log2_weight_denom, + h->chroma_weight[refn0][0][0][0], + h->chroma_weight[refn1][1][0][0], + h->chroma_weight[refn0][0][0][1] + + h->chroma_weight[refn1][1][0][1]); + chroma_weight_avg(dest_cr, tmp_cr, h->mb_uvlinesize, chroma_height, + h->chroma_log2_weight_denom, + h->chroma_weight[refn0][0][1][0], + h->chroma_weight[refn1][1][1][0], + h->chroma_weight[refn0][0][1][1] + + h->chroma_weight[refn1][1][1][1]); } - }else{ - int list = list1 ? 1 : 0; - int refn = h->ref_cache[list][ scan8[n] ]; - Picture *ref= &h->ref_list[list][refn]; + } else { + int list = list1 ? 1 : 0; + int refn = h->ref_cache[list][scan8[n]]; + Picture *ref = &h->ref_list[list][refn]; mc_dir_part(h, ref, n, square, height, delta, list, dest_y, dest_cb, dest_cr, x_offset, y_offset, qpix_put, chroma_put, pixel_shift, chroma_idc); - luma_weight_op(dest_y, h->mb_linesize, height, h->luma_log2_weight_denom, - h->luma_weight[refn][list][0], h->luma_weight[refn][list][1]); - if(h->use_weight_chroma){ - chroma_weight_op(dest_cb, h->mb_uvlinesize, chroma_height, h->chroma_log2_weight_denom, - h->chroma_weight[refn][list][0][0], h->chroma_weight[refn][list][0][1]); - chroma_weight_op(dest_cr, h->mb_uvlinesize, chroma_height, h->chroma_log2_weight_denom, - h->chroma_weight[refn][list][1][0], h->chroma_weight[refn][list][1][1]); + luma_weight_op(dest_y, h->mb_linesize, height, + h->luma_log2_weight_denom, + h->luma_weight[refn][list][0], + h->luma_weight[refn][list][1]); + if (h->use_weight_chroma) { + chroma_weight_op(dest_cb, h->mb_uvlinesize, chroma_height, + h->chroma_log2_weight_denom, + h->chroma_weight[refn][list][0][0], + h->chroma_weight[refn][list][0][1]); + chroma_weight_op(dest_cr, h->mb_uvlinesize, chroma_height, + h->chroma_log2_weight_denom, + h->chroma_weight[refn][list][1][0], + h->chroma_weight[refn][list][1][1]); } } } -static av_always_inline void -mc_part(H264Context *h, int n, int square, int height, int delta, - uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, - int x_offset, int y_offset, - qpel_mc_func *qpix_put, h264_chroma_mc_func chroma_put, - qpel_mc_func *qpix_avg, h264_chroma_mc_func chroma_avg, - h264_weight_func *weight_op, h264_biweight_func *weight_avg, - int list0, int list1, int pixel_shift, int chroma_idc) +static av_always_inline void mc_part(H264Context *h, int n, int square, + int height, int delta, + uint8_t *dest_y, uint8_t *dest_cb, + uint8_t *dest_cr, + int x_offset, int y_offset, + qpel_mc_func *qpix_put, + h264_chroma_mc_func chroma_put, + qpel_mc_func *qpix_avg, + h264_chroma_mc_func chroma_avg, + h264_weight_func *weight_op, + h264_biweight_func *weight_avg, + int list0, int list1, + int pixel_shift, int chroma_idc) { - if((h->use_weight==2 && list0 && list1 - && (h->implicit_weight[ h->ref_cache[0][scan8[n]] ][ h->ref_cache[1][scan8[n]] ][h->s.mb_y&1] != 32)) - || h->use_weight==1) + if ((h->use_weight == 2 && list0 && list1 && + (h->implicit_weight[h->ref_cache[0][scan8[n]]][h->ref_cache[1][scan8[n]]][h->s.mb_y & 1] != 32)) || + h->use_weight == 1) mc_part_weighted(h, n, square, height, delta, dest_y, dest_cb, dest_cr, x_offset, y_offset, qpix_put, chroma_put, weight_op[0], weight_op[1], weight_avg[0], @@ -692,52 +739,60 @@ mc_part(H264Context *h, int n, int square, int height, int delta, chroma_avg, list0, list1, pixel_shift, chroma_idc); } -static av_always_inline void -prefetch_motion(H264Context *h, int list, int pixel_shift, int chroma_idc) +static av_always_inline void prefetch_motion(H264Context *h, int list, + int pixel_shift, int chroma_idc) { /* fetch pixels for estimated mv 4 macroblocks ahead * optimized for 64byte cache lines */ - MpegEncContext * const s = &h->s; + MpegEncContext *const s = &h->s; const int refn = h->ref_cache[list][scan8[0]]; - if(refn >= 0){ - const int mx= (h->mv_cache[list][scan8[0]][0]>>2) + 16*s->mb_x + 8; - const int my= (h->mv_cache[list][scan8[0]][1]>>2) + 16*s->mb_y; + if (refn >= 0) { + const int mx = (h->mv_cache[list][scan8[0]][0] >> 2) + 16 * s->mb_x + 8; + const int my = (h->mv_cache[list][scan8[0]][1] >> 2) + 16 * s->mb_y; uint8_t **src = h->ref_list[list][refn].f.data; - int off= (mx << pixel_shift) + (my + (s->mb_x&3)*4)*h->mb_linesize + (64 << pixel_shift); - s->dsp.prefetch(src[0]+off, s->linesize, 4); + int off = (mx << pixel_shift) + + (my + (s->mb_x & 3) * 4) * h->mb_linesize + + (64 << pixel_shift); + s->dsp.prefetch(src[0] + off, s->linesize, 4); if (chroma_idc == 3 /* yuv444 */) { - 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); - s->dsp.prefetch(src[1]+off, src[2]-src[1], 2); + 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); + s->dsp.prefetch(src[1] + off, src[2] - src[1], 2); } } } -static av_always_inline void hl_motion(H264Context *h, uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, - qpel_mc_func (*qpix_put)[16], h264_chroma_mc_func (*chroma_put), - qpel_mc_func (*qpix_avg)[16], h264_chroma_mc_func (*chroma_avg), - h264_weight_func *weight_op, h264_biweight_func *weight_avg, - int pixel_shift, int chroma_idc) +static av_always_inline void hl_motion(H264Context *h, uint8_t *dest_y, + uint8_t *dest_cb, uint8_t *dest_cr, + qpel_mc_func(*qpix_put)[16], + h264_chroma_mc_func(*chroma_put), + qpel_mc_func(*qpix_avg)[16], + h264_chroma_mc_func(*chroma_avg), + h264_weight_func *weight_op, + h264_biweight_func *weight_avg, + int pixel_shift, int chroma_idc) { - MpegEncContext * const s = &h->s; - const int mb_xy= h->mb_xy; + MpegEncContext *const s = &h->s; + const int mb_xy = h->mb_xy; const int mb_type = s->current_picture.f.mb_type[mb_xy]; assert(IS_INTER(mb_type)); - if(HAVE_THREADS && (s->avctx->active_thread_type & FF_THREAD_FRAME)) + if (HAVE_THREADS && (s->avctx->active_thread_type & FF_THREAD_FRAME)) await_references(h); prefetch_motion(h, 0, pixel_shift, chroma_idc); - if(IS_16X16(mb_type)){ + if (IS_16X16(mb_type)) { mc_part(h, 0, 1, 16, 0, dest_y, dest_cb, dest_cr, 0, 0, qpix_put[0], chroma_put[0], qpix_avg[0], chroma_avg[0], weight_op, weight_avg, IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1), pixel_shift, chroma_idc); - }else if(IS_16X8(mb_type)){ + } else if (IS_16X8(mb_type)) { mc_part(h, 0, 0, 8, 8 << pixel_shift, dest_y, dest_cb, dest_cr, 0, 0, qpix_put[1], chroma_put[0], qpix_avg[1], chroma_avg[0], weight_op, weight_avg, @@ -748,67 +803,73 @@ static av_always_inline void hl_motion(H264Context *h, uint8_t *dest_y, uint8_t weight_op, weight_avg, IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1), pixel_shift, chroma_idc); - }else if(IS_8X16(mb_type)){ - mc_part(h, 0, 0, 16, 8*h->mb_linesize, dest_y, dest_cb, dest_cr, 0, 0, + } else if (IS_8X16(mb_type)) { + mc_part(h, 0, 0, 16, 8 * h->mb_linesize, dest_y, dest_cb, dest_cr, 0, 0, qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1], &weight_op[1], &weight_avg[1], IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1), pixel_shift, chroma_idc); - mc_part(h, 4, 0, 16, 8*h->mb_linesize, dest_y, dest_cb, dest_cr, 4, 0, + mc_part(h, 4, 0, 16, 8 * h->mb_linesize, dest_y, dest_cb, dest_cr, 4, 0, qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1], &weight_op[1], &weight_avg[1], IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1), pixel_shift, chroma_idc); - }else{ + } else { int i; assert(IS_8X8(mb_type)); - for(i=0; i<4; i++){ - const int sub_mb_type= h->sub_mb_type[i]; - const int n= 4*i; - int x_offset= (i&1)<<2; - int y_offset= (i&2)<<1; + for (i = 0; i < 4; i++) { + const int sub_mb_type = h->sub_mb_type[i]; + const int n = 4 * i; + int x_offset = (i & 1) << 2; + int y_offset = (i & 2) << 1; - if(IS_SUB_8X8(sub_mb_type)){ - mc_part(h, n, 1, 8, 0, dest_y, dest_cb, dest_cr, x_offset, y_offset, - qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1], - &weight_op[1], &weight_avg[1], - IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), - pixel_shift, chroma_idc); - }else if(IS_SUB_8X4(sub_mb_type)){ - mc_part(h, n , 0, 4, 4 << pixel_shift, dest_y, dest_cb, dest_cr, x_offset, y_offset, - qpix_put[2], chroma_put[1], qpix_avg[2], chroma_avg[1], - &weight_op[1], &weight_avg[1], - IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), - pixel_shift, chroma_idc); - mc_part(h, n+2, 0, 4, 4 << pixel_shift, dest_y, dest_cb, dest_cr, x_offset, y_offset+2, - qpix_put[2], chroma_put[1], qpix_avg[2], chroma_avg[1], - &weight_op[1], &weight_avg[1], - IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), - pixel_shift, chroma_idc); - }else if(IS_SUB_4X8(sub_mb_type)){ - mc_part(h, n , 0, 8, 4*h->mb_linesize, dest_y, dest_cb, dest_cr, x_offset, y_offset, - qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2], - &weight_op[2], &weight_avg[2], - IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), - pixel_shift, chroma_idc); - mc_part(h, n+1, 0, 8, 4*h->mb_linesize, dest_y, dest_cb, dest_cr, x_offset+2, y_offset, - qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2], - &weight_op[2], &weight_avg[2], - IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), - pixel_shift, chroma_idc); - }else{ - int j; - assert(IS_SUB_4X4(sub_mb_type)); - for(j=0; j<4; j++){ - int sub_x_offset= x_offset + 2*(j&1); - int sub_y_offset= y_offset + (j&2); - mc_part(h, n+j, 1, 4, 0, dest_y, dest_cb, dest_cr, sub_x_offset, sub_y_offset, + if (IS_SUB_8X8(sub_mb_type)) { + mc_part(h, n, 1, 8, 0, dest_y, dest_cb, dest_cr, + x_offset, y_offset, + qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1], + &weight_op[1], &weight_avg[1], + IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), + pixel_shift, chroma_idc); + } else if (IS_SUB_8X4(sub_mb_type)) { + mc_part(h, n, 0, 4, 4 << pixel_shift, dest_y, dest_cb, dest_cr, + x_offset, y_offset, + qpix_put[2], chroma_put[1], qpix_avg[2], chroma_avg[1], + &weight_op[1], &weight_avg[1], + IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), + pixel_shift, chroma_idc); + mc_part(h, n + 2, 0, 4, 4 << pixel_shift, + dest_y, dest_cb, dest_cr, x_offset, y_offset + 2, + qpix_put[2], chroma_put[1], qpix_avg[2], chroma_avg[1], + &weight_op[1], &weight_avg[1], + IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), + pixel_shift, chroma_idc); + } else if (IS_SUB_4X8(sub_mb_type)) { + mc_part(h, n, 0, 8, 4 * h->mb_linesize, + dest_y, dest_cb, dest_cr, x_offset, y_offset, qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2], &weight_op[2], &weight_avg[2], IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), pixel_shift, chroma_idc); + mc_part(h, n + 1, 0, 8, 4 * h->mb_linesize, + dest_y, dest_cb, dest_cr, x_offset + 2, y_offset, + qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2], + &weight_op[2], &weight_avg[2], + IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), + pixel_shift, chroma_idc); + } else { + int j; + assert(IS_SUB_4X4(sub_mb_type)); + for (j = 0; j < 4; j++) { + int sub_x_offset = x_offset + 2 * (j & 1); + int sub_y_offset = y_offset + (j & 2); + mc_part(h, n + j, 1, 4, 0, + dest_y, dest_cb, dest_cr, sub_x_offset, sub_y_offset, + qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2], + &weight_op[2], &weight_avg[2], + IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), + pixel_shift, chroma_idc); } } } @@ -817,29 +878,36 @@ static av_always_inline void hl_motion(H264Context *h, uint8_t *dest_y, uint8_t prefetch_motion(h, 1, pixel_shift, chroma_idc); } -static av_always_inline void -hl_motion_420(H264Context *h, uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, - qpel_mc_func (*qpix_put)[16], h264_chroma_mc_func (*chroma_put), - qpel_mc_func (*qpix_avg)[16], h264_chroma_mc_func (*chroma_avg), - h264_weight_func *weight_op, h264_biweight_func *weight_avg, - int pixel_shift) +static av_always_inline void hl_motion_420(H264Context *h, uint8_t *dest_y, + uint8_t *dest_cb, uint8_t *dest_cr, + qpel_mc_func(*qpix_put)[16], + h264_chroma_mc_func(*chroma_put), + qpel_mc_func(*qpix_avg)[16], + h264_chroma_mc_func(*chroma_avg), + h264_weight_func *weight_op, + h264_biweight_func *weight_avg, + int pixel_shift) { hl_motion(h, dest_y, dest_cb, dest_cr, qpix_put, chroma_put, qpix_avg, chroma_avg, weight_op, weight_avg, pixel_shift, 1); } -static av_always_inline void -hl_motion_422(H264Context *h, uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, - qpel_mc_func (*qpix_put)[16], h264_chroma_mc_func (*chroma_put), - qpel_mc_func (*qpix_avg)[16], h264_chroma_mc_func (*chroma_avg), - h264_weight_func *weight_op, h264_biweight_func *weight_avg, - int pixel_shift) +static av_always_inline void hl_motion_422(H264Context *h, uint8_t *dest_y, + uint8_t *dest_cb, uint8_t *dest_cr, + qpel_mc_func(*qpix_put)[16], + h264_chroma_mc_func(*chroma_put), + qpel_mc_func(*qpix_avg)[16], + h264_chroma_mc_func(*chroma_avg), + h264_weight_func *weight_op, + h264_biweight_func *weight_avg, + int pixel_shift) { hl_motion(h, dest_y, dest_cb, dest_cr, qpix_put, chroma_put, qpix_avg, chroma_avg, weight_op, weight_avg, pixel_shift, 2); } -static void free_tables(H264Context *h, int free_rbsp){ +static void free_tables(H264Context *h, int free_rbsp) +{ int i; H264Context *hx; @@ -851,135 +919,150 @@ static void free_tables(H264Context *h, int free_rbsp){ av_freep(&h->direct_table); av_freep(&h->non_zero_count); av_freep(&h->slice_table_base); - h->slice_table= NULL; + h->slice_table = NULL; av_freep(&h->list_counts); av_freep(&h->mb2b_xy); av_freep(&h->mb2br_xy); - for(i = 0; i < MAX_THREADS; i++) { + for (i = 0; i < MAX_THREADS; i++) { hx = h->thread_context[i]; - if(!hx) continue; + if (!hx) + continue; av_freep(&hx->top_borders[1]); av_freep(&hx->top_borders[0]); av_freep(&hx->s.obmc_scratchpad); - if (free_rbsp){ + if (free_rbsp) { av_freep(&hx->rbsp_buffer[1]); av_freep(&hx->rbsp_buffer[0]); hx->rbsp_buffer_size[0] = 0; hx->rbsp_buffer_size[1] = 0; } - if (i) av_freep(&h->thread_context[i]); + if (i) + av_freep(&h->thread_context[i]); } } -static void init_dequant8_coeff_table(H264Context *h){ - int i,j,q,x; - const int max_qp = 51 + 6*(h->sps.bit_depth_luma-8); +static void init_dequant8_coeff_table(H264Context *h) +{ + int i, j, q, x; + const int max_qp = 51 + 6 * (h->sps.bit_depth_luma - 8); - for(i=0; i<6; i++ ){ + for (i = 0; i < 6; i++) { h->dequant8_coeff[i] = h->dequant8_buffer[i]; - for(j=0; jpps.scaling_matrix8[j], h->pps.scaling_matrix8[i], 64*sizeof(uint8_t))){ + for (j = 0; j < i; j++) + if (!memcmp(h->pps.scaling_matrix8[j], h->pps.scaling_matrix8[i], + 64 * sizeof(uint8_t))) { h->dequant8_coeff[i] = h->dequant8_buffer[j]; break; } - } - if(jdequant8_coeff[i][q][(x>>3)|((x&7)<<3)] = - ((uint32_t)dequant8_coeff_init[idx][ dequant8_coeff_init_scan[((x>>1)&12) | (x&3)] ] * - h->pps.scaling_matrix8[i][x]) << shift; + int idx = rem6[q]; + for (x = 0; x < 64; x++) + h->dequant8_coeff[i][q][(x >> 3) | ((x & 7) << 3)] = + ((uint32_t)dequant8_coeff_init[idx][dequant8_coeff_init_scan[((x >> 1) & 12) | (x & 3)]] * + h->pps.scaling_matrix8[i][x]) << shift; } } } -static void init_dequant4_coeff_table(H264Context *h){ - int i,j,q,x; - const int max_qp = 51 + 6*(h->sps.bit_depth_luma-8); - for(i=0; i<6; i++ ){ +static void init_dequant4_coeff_table(H264Context *h) +{ + int i, j, q, x; + const int max_qp = 51 + 6 * (h->sps.bit_depth_luma - 8); + for (i = 0; i < 6; i++) { h->dequant4_coeff[i] = h->dequant4_buffer[i]; - for(j=0; jpps.scaling_matrix4[j], h->pps.scaling_matrix4[i], 16*sizeof(uint8_t))){ + for (j = 0; j < i; j++) + if (!memcmp(h->pps.scaling_matrix4[j], h->pps.scaling_matrix4[i], + 16 * sizeof(uint8_t))) { h->dequant4_coeff[i] = h->dequant4_buffer[j]; break; } - } - if(jdequant4_coeff[i][q][(x>>2)|((x<<2)&0xF)] = - ((uint32_t)dequant4_coeff_init[idx][(x&1) + ((x>>2)&1)] * - h->pps.scaling_matrix4[i][x]) << shift; + int idx = rem6[q]; + for (x = 0; x < 16; x++) + h->dequant4_coeff[i][q][(x >> 2) | ((x << 2) & 0xF)] = + ((uint32_t)dequant4_coeff_init[idx][(x & 1) + ((x >> 2) & 1)] * + h->pps.scaling_matrix4[i][x]) << shift; } } } -static void init_dequant_tables(H264Context *h){ - int i,x; +static void init_dequant_tables(H264Context *h) +{ + int i, x; init_dequant4_coeff_table(h); - if(h->pps.transform_8x8_mode) + if (h->pps.transform_8x8_mode) init_dequant8_coeff_table(h); - if(h->sps.transform_bypass){ - for(i=0; i<6; i++) - for(x=0; x<16; x++) - h->dequant4_coeff[i][0][x] = 1<<6; - if(h->pps.transform_8x8_mode) - for(i=0; i<6; i++) - for(x=0; x<64; x++) - h->dequant8_coeff[i][0][x] = 1<<6; + if (h->sps.transform_bypass) { + for (i = 0; i < 6; i++) + for (x = 0; x < 16; x++) + h->dequant4_coeff[i][0][x] = 1 << 6; + if (h->pps.transform_8x8_mode) + for (i = 0; i < 6; i++) + for (x = 0; x < 64; x++) + h->dequant8_coeff[i][0][x] = 1 << 6; } } +int ff_h264_alloc_tables(H264Context *h) +{ + MpegEncContext *const s = &h->s; + const int big_mb_num = s->mb_stride * (s->mb_height + 1); + const int row_mb_num = s->mb_stride * 2 * s->avctx->thread_count; + int x, y; -int ff_h264_alloc_tables(H264Context *h){ - MpegEncContext * const s = &h->s; - const int big_mb_num= s->mb_stride * (s->mb_height+1); - const int row_mb_num= 2*s->mb_stride*s->avctx->thread_count; - int x,y; + FF_ALLOCZ_OR_GOTO(h->s.avctx, h->intra4x4_pred_mode, + row_mb_num * 8 * sizeof(uint8_t), fail) + FF_ALLOCZ_OR_GOTO(h->s.avctx, h->non_zero_count, + big_mb_num * 48 * sizeof(uint8_t), fail) + FF_ALLOCZ_OR_GOTO(h->s.avctx, h->slice_table_base, + (big_mb_num + s->mb_stride) * sizeof(*h->slice_table_base), fail) + FF_ALLOCZ_OR_GOTO(h->s.avctx, h->cbp_table, + big_mb_num * sizeof(uint16_t), fail) + FF_ALLOCZ_OR_GOTO(h->s.avctx, h->chroma_pred_mode_table, + big_mb_num * sizeof(uint8_t), fail) + FF_ALLOCZ_OR_GOTO(h->s.avctx, h->mvd_table[0], + 16 * row_mb_num * sizeof(uint8_t), fail); + FF_ALLOCZ_OR_GOTO(h->s.avctx, h->mvd_table[1], + 16 * row_mb_num * sizeof(uint8_t), fail); + FF_ALLOCZ_OR_GOTO(h->s.avctx, h->direct_table, + 4 * big_mb_num * sizeof(uint8_t), fail); + FF_ALLOCZ_OR_GOTO(h->s.avctx, h->list_counts, + big_mb_num * sizeof(uint8_t), fail) - FF_ALLOCZ_OR_GOTO(h->s.avctx, h->intra4x4_pred_mode, row_mb_num * 8 * sizeof(uint8_t), fail) + memset(h->slice_table_base, -1, + (big_mb_num + s->mb_stride) * sizeof(*h->slice_table_base)); + h->slice_table = h->slice_table_base + s->mb_stride * 2 + 1; - FF_ALLOCZ_OR_GOTO(h->s.avctx, h->non_zero_count , big_mb_num * 48 * sizeof(uint8_t), fail) - FF_ALLOCZ_OR_GOTO(h->s.avctx, h->slice_table_base , (big_mb_num+s->mb_stride) * sizeof(*h->slice_table_base), fail) - FF_ALLOCZ_OR_GOTO(h->s.avctx, h->cbp_table, big_mb_num * sizeof(uint16_t), fail) + FF_ALLOCZ_OR_GOTO(h->s.avctx, h->mb2b_xy, + big_mb_num * sizeof(uint32_t), fail); + FF_ALLOCZ_OR_GOTO(h->s.avctx, h->mb2br_xy, + big_mb_num * sizeof(uint32_t), fail); + for (y = 0; y < s->mb_height; y++) + for (x = 0; x < s->mb_width; x++) { + const int mb_xy = x + y * s->mb_stride; + const int b_xy = 4 * x + 4 * y * h->b_stride; - FF_ALLOCZ_OR_GOTO(h->s.avctx, h->chroma_pred_mode_table, big_mb_num * sizeof(uint8_t), fail) - FF_ALLOCZ_OR_GOTO(h->s.avctx, h->mvd_table[0], 16*row_mb_num * sizeof(uint8_t), fail); - FF_ALLOCZ_OR_GOTO(h->s.avctx, h->mvd_table[1], 16*row_mb_num * sizeof(uint8_t), fail); - FF_ALLOCZ_OR_GOTO(h->s.avctx, h->direct_table, 4*big_mb_num * sizeof(uint8_t) , fail); - FF_ALLOCZ_OR_GOTO(h->s.avctx, h->list_counts, big_mb_num * sizeof(uint8_t), fail) - - memset(h->slice_table_base, -1, (big_mb_num+s->mb_stride) * sizeof(*h->slice_table_base)); - h->slice_table= h->slice_table_base + s->mb_stride*2 + 1; - - FF_ALLOCZ_OR_GOTO(h->s.avctx, h->mb2b_xy , big_mb_num * sizeof(uint32_t), fail); - FF_ALLOCZ_OR_GOTO(h->s.avctx, h->mb2br_xy , big_mb_num * sizeof(uint32_t), fail); - for(y=0; ymb_height; y++){ - for(x=0; xmb_width; x++){ - const int mb_xy= x + y*s->mb_stride; - const int b_xy = 4*x + 4*y*h->b_stride; - - h->mb2b_xy [mb_xy]= b_xy; - h->mb2br_xy[mb_xy]= 8*(FMO ? mb_xy : (mb_xy % (2*s->mb_stride))); + h->mb2b_xy[mb_xy] = b_xy; + h->mb2br_xy[mb_xy] = 8 * (FMO ? mb_xy : (mb_xy % (2 * s->mb_stride))); } - } s->obmc_scratchpad = NULL; - if(!h->dequant4_coeff[0]) + if (!h->dequant4_coeff[0]) init_dequant_tables(h); return 0; + fail: free_tables(h, 1); return -1; @@ -988,87 +1071,99 @@ fail: /** * Mimic alloc_tables(), but for every context thread. */ -static void clone_tables(H264Context *dst, H264Context *src, int i){ - MpegEncContext * const s = &src->s; - dst->intra4x4_pred_mode = src->intra4x4_pred_mode + i*8*2*s->mb_stride; - dst->non_zero_count = src->non_zero_count; - dst->slice_table = src->slice_table; - dst->cbp_table = src->cbp_table; - dst->mb2b_xy = src->mb2b_xy; - dst->mb2br_xy = src->mb2br_xy; - dst->chroma_pred_mode_table = src->chroma_pred_mode_table; - dst->mvd_table[0] = src->mvd_table[0] + i*8*2*s->mb_stride; - dst->mvd_table[1] = src->mvd_table[1] + i*8*2*s->mb_stride; - dst->direct_table = src->direct_table; - dst->list_counts = src->list_counts; - - dst->s.obmc_scratchpad = NULL; - ff_h264_pred_init(&dst->hpc, src->s.codec_id, src->sps.bit_depth_luma, src->sps.chroma_format_idc); +static void clone_tables(H264Context *dst, H264Context *src, int i) +{ + MpegEncContext *const s = &src->s; + dst->intra4x4_pred_mode = src->intra4x4_pred_mode + i * 8 * 2 * s->mb_stride; + dst->non_zero_count = src->non_zero_count; + dst->slice_table = src->slice_table; + dst->cbp_table = src->cbp_table; + dst->mb2b_xy = src->mb2b_xy; + dst->mb2br_xy = src->mb2br_xy; + dst->chroma_pred_mode_table = src->chroma_pred_mode_table; + dst->mvd_table[0] = src->mvd_table[0] + i * 8 * 2 * s->mb_stride; + dst->mvd_table[1] = src->mvd_table[1] + i * 8 * 2 * s->mb_stride; + dst->direct_table = src->direct_table; + dst->list_counts = src->list_counts; + dst->s.obmc_scratchpad = NULL; + ff_h264_pred_init(&dst->hpc, src->s.codec_id, src->sps.bit_depth_luma, + src->sps.chroma_format_idc); } /** * Init context * Allocate buffers which are not shared amongst multiple threads. */ -static int context_init(H264Context *h){ - FF_ALLOCZ_OR_GOTO(h->s.avctx, h->top_borders[0], h->s.mb_width * 16*3 * sizeof(uint8_t)*2, fail) - FF_ALLOCZ_OR_GOTO(h->s.avctx, h->top_borders[1], h->s.mb_width * 16*3 * sizeof(uint8_t)*2, fail) +static int context_init(H264Context *h) +{ + FF_ALLOCZ_OR_GOTO(h->s.avctx, h->top_borders[0], + h->s.mb_width * 16 * 3 * sizeof(uint8_t) * 2, fail) + FF_ALLOCZ_OR_GOTO(h->s.avctx, h->top_borders[1], + h->s.mb_width * 16 * 3 * sizeof(uint8_t) * 2, fail) - h->ref_cache[0][scan8[5 ]+1] = h->ref_cache[0][scan8[7 ]+1] = h->ref_cache[0][scan8[13]+1] = - h->ref_cache[1][scan8[5 ]+1] = h->ref_cache[1][scan8[7 ]+1] = h->ref_cache[1][scan8[13]+1] = PART_NOT_AVAILABLE; + h->ref_cache[0][scan8[5] + 1] = + h->ref_cache[0][scan8[7] + 1] = + h->ref_cache[0][scan8[13] + 1] = + h->ref_cache[1][scan8[5] + 1] = + h->ref_cache[1][scan8[7] + 1] = + h->ref_cache[1][scan8[13] + 1] = PART_NOT_AVAILABLE; return 0; + fail: return -1; // free_tables will clean up for us } static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size); -static av_cold void common_init(H264Context *h){ - MpegEncContext * const s = &h->s; +static av_cold void common_init(H264Context *h) +{ + MpegEncContext *const s = &h->s; - s->width = s->avctx->width; - s->height = s->avctx->height; - s->codec_id= s->avctx->codec->id; + s->width = s->avctx->width; + 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); - h->dequant_coeff_pps= -1; - s->unrestricted_mv=1; + h->dequant_coeff_pps = -1; + s->unrestricted_mv = 1; - ff_dsputil_init(&s->dsp, s->avctx); // needed so that idct permutation is known early + /* needed so that IDCT permutation is known early */ + ff_dsputil_init(&s->dsp, s->avctx); - memset(h->pps.scaling_matrix4, 16, 6*16*sizeof(uint8_t)); - memset(h->pps.scaling_matrix8, 16, 2*64*sizeof(uint8_t)); + 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) { AVCodecContext *avctx = h->s.avctx; - if(avctx->extradata[0] == 1){ + if (avctx->extradata[0] == 1) { int i, cnt, nalsize; unsigned char *p = avctx->extradata; h->is_avc = 1; - if(avctx->extradata_size < 7) { + if (avctx->extradata_size < 7) { av_log(avctx, AV_LOG_ERROR, "avcC too short\n"); return -1; } /* sps and pps in the avcC always have length coded with 2 bytes, - so put a fake nal_length_size = 2 while parsing them */ + * so put a fake nal_length_size = 2 while parsing them */ h->nal_length_size = 2; // Decode sps from avcC - cnt = *(p+5) & 0x1f; // Number of sps - p += 6; + cnt = *(p + 5) & 0x1f; // Number of sps + p += 6; for (i = 0; i < cnt; i++) { nalsize = AV_RB16(p) + 2; if (p - avctx->extradata + nalsize > avctx->extradata_size) return -1; - if(decode_nal_units(h, p, nalsize) < 0) { - av_log(avctx, AV_LOG_ERROR, "Decoding sps %d from avcC failed\n", i); + if (decode_nal_units(h, p, nalsize) < 0) { + av_log(avctx, AV_LOG_ERROR, + "Decoding sps %d from avcC failed\n", i); return -1; } p += nalsize; @@ -1080,24 +1175,26 @@ int ff_h264_decode_extradata(H264Context *h) if (p - avctx->extradata + nalsize > avctx->extradata_size) return -1; if (decode_nal_units(h, p, nalsize) < 0) { - av_log(avctx, AV_LOG_ERROR, "Decoding pps %d from avcC failed\n", i); + av_log(avctx, AV_LOG_ERROR, + "Decoding pps %d from avcC failed\n", i); return -1; } p += nalsize; } - // Now store right nal length size, that will be use to parse all other nals + // Now store right nal length size, that will be used to parse all other nals h->nal_length_size = (avctx->extradata[4] & 0x03) + 1; } else { h->is_avc = 0; - if(decode_nal_units(h, avctx->extradata, avctx->extradata_size) < 0) + if (decode_nal_units(h, avctx->extradata, avctx->extradata_size) < 0) return -1; } return 0; } -av_cold int ff_h264_decode_init(AVCodecContext *avctx){ - H264Context *h= avctx->priv_data; - MpegEncContext * const s = &h->s; +av_cold int ff_h264_decode_init(AVCodecContext *avctx) +{ + H264Context *h = avctx->priv_data; + MpegEncContext *const s = &h->s; int i; ff_MPV_decode_defaults(s); @@ -1105,14 +1202,14 @@ av_cold int ff_h264_decode_init(AVCodecContext *avctx){ s->avctx = avctx; common_init(h); - s->out_format = FMT_H264; - s->workaround_bugs= avctx->workaround_bugs; + s->out_format = FMT_H264; + s->workaround_bugs = avctx->workaround_bugs; - // set defaults -// s->decode_mb= ff_h263_decode_mb; + /* set defaults */ + // s->decode_mb = ff_h263_decode_mb; s->quarter_sample = 1; - if(!avctx->has_b_frames) - s->low_delay= 1; + if (!avctx->has_b_frames) + s->low_delay = 1; avctx->chroma_sample_location = AVCHROMA_LOC_LEFT; @@ -1122,39 +1219,43 @@ av_cold int ff_h264_decode_init(AVCodecContext *avctx){ h->sps.bit_depth_luma = avctx->bits_per_raw_sample = 8; h->thread_context[0] = h; - h->outputed_poc = h->next_outputed_poc = INT_MIN; + h->outputed_poc = h->next_outputed_poc = INT_MIN; for (i = 0; i < MAX_DELAYED_PIC_COUNT; i++) h->last_pocs[i] = INT_MIN; - h->prev_poc_msb= 1<<16; - h->x264_build = -1; + h->prev_poc_msb = 1 << 16; + h->x264_build = -1; ff_h264_reset_sei(h); - if(avctx->codec_id == CODEC_ID_H264){ - if(avctx->ticks_per_frame == 1){ - s->avctx->time_base.den *=2; - } + if (avctx->codec_id == CODEC_ID_H264) { + if (avctx->ticks_per_frame == 1) + s->avctx->time_base.den *= 2; avctx->ticks_per_frame = 2; } - if(avctx->extradata_size > 0 && avctx->extradata && + if (avctx->extradata_size > 0 && avctx->extradata && ff_h264_decode_extradata(h)) return -1; - if(h->sps.bitstream_restriction_flag && s->avctx->has_b_frames < h->sps.num_reorder_frames){ + if (h->sps.bitstream_restriction_flag && + s->avctx->has_b_frames < h->sps.num_reorder_frames) { s->avctx->has_b_frames = h->sps.num_reorder_frames; - s->low_delay = 0; + s->low_delay = 0; } return 0; } -#define IN_RANGE(a, b, size) (((a) >= (b)) && ((a) < ((b)+(size)))) -static void copy_picture_range(Picture **to, Picture **from, int count, MpegEncContext *new_base, MpegEncContext *old_base) +#define IN_RANGE(a, b, size) (((a) >= (b)) && ((a) < ((b) + (size)))) + +static void copy_picture_range(Picture **to, Picture **from, int count, + MpegEncContext *new_base, + MpegEncContext *old_base) { int i; - for (i=0; ipicture, sizeof(Picture) * old_base->picture_count) || + IN_RANGE(from[i], old_base->picture, + sizeof(Picture) * old_base->picture_count) || !from[i])); to[i] = REBASE_PICTURE(from[i], new_base, old_base); } @@ -1164,16 +1265,20 @@ static void copy_parameter_set(void **to, void **from, int count, int size) { int i; - for (i=0; ipriv_data; +static int decode_init_thread_copy(AVCodecContext *avctx) +{ + H264Context *h = avctx->priv_data; if (!avctx->internal->is_copy) return 0; @@ -1183,27 +1288,36 @@ static int decode_init_thread_copy(AVCodecContext *avctx){ return 0; } -#define copy_fields(to, from, start_field, end_field) memcpy(&to->start_field, &from->start_field, (char*)&to->end_field - (char*)&to->start_field) -static int decode_update_thread_context(AVCodecContext *dst, const AVCodecContext *src){ - H264Context *h= dst->priv_data, *h1= src->priv_data; - MpegEncContext * const s = &h->s, * const s1 = &h1->s; +#define copy_fields(to, from, start_field, end_field) \ + memcpy(&to->start_field, &from->start_field, \ + (char *)&to->end_field - (char *)&to->start_field) + +static int decode_update_thread_context(AVCodecContext *dst, + const AVCodecContext *src) +{ + H264Context *h = dst->priv_data, *h1 = src->priv_data; + MpegEncContext *const s = &h->s, *const s1 = &h1->s; int inited = s->context_initialized, err; int i; - if(dst == src || !s1->context_initialized) return 0; + if (dst == src || !s1->context_initialized) + return 0; err = ff_mpeg_update_thread_context(dst, src); - if(err) return err; + if (err) + return err; - //FIXME handle width/height changing - if(!inited){ - for(i = 0; i < MAX_SPS_COUNT; i++) + // FIXME handle width/height changing + if (!inited) { + for (i = 0; i < MAX_SPS_COUNT; i++) av_freep(h->sps_buffers + i); - for(i = 0; i < MAX_PPS_COUNT; i++) + for (i = 0; i < MAX_PPS_COUNT; i++) av_freep(h->pps_buffers + i); - memcpy(&h->s + 1, &h1->s + 1, sizeof(H264Context) - sizeof(MpegEncContext)); //copy all fields after MpegEnc + // copy all fields after MpegEnc + memcpy(&h->s + 1, &h1->s + 1, + sizeof(H264Context) - sizeof(MpegEncContext)); memset(h->sps_buffers, 0, sizeof(h->sps_buffers)); memset(h->pps_buffers, 0, sizeof(h->pps_buffers)); if (ff_h264_alloc_tables(h) < 0) { @@ -1212,76 +1326,83 @@ static int decode_update_thread_context(AVCodecContext *dst, const AVCodecContex } context_init(h); - for(i=0; i<2; i++){ - h->rbsp_buffer[i] = NULL; + for (i = 0; i < 2; i++) { + h->rbsp_buffer[i] = NULL; h->rbsp_buffer_size[i] = 0; } h->thread_context[0] = h; - // frame_start may not be called for the next thread (if it's decoding a bottom field) - // so this has to be allocated here - h->s.obmc_scratchpad = av_malloc(16*6*s->linesize); + /* frame_start may not be called for the next thread (if it's decoding + * a bottom field) so this has to be allocated here */ + h->s.obmc_scratchpad = av_malloc(16 * 6 * s->linesize); s->dsp.clear_blocks(h->mb); - s->dsp.clear_blocks(h->mb+(24*16<pixel_shift)); + s->dsp.clear_blocks(h->mb + (24 * 16 << h->pixel_shift)); } - //extradata/NAL handling - h->is_avc = h1->is_avc; + // extradata/NAL handling + h->is_avc = h1->is_avc; - //SPS/PPS - copy_parameter_set((void**)h->sps_buffers, (void**)h1->sps_buffers, MAX_SPS_COUNT, sizeof(SPS)); - h->sps = h1->sps; - copy_parameter_set((void**)h->pps_buffers, (void**)h1->pps_buffers, MAX_PPS_COUNT, sizeof(PPS)); - h->pps = h1->pps; + // SPS/PPS + copy_parameter_set((void **)h->sps_buffers, (void **)h1->sps_buffers, + MAX_SPS_COUNT, sizeof(SPS)); + h->sps = h1->sps; + copy_parameter_set((void **)h->pps_buffers, (void **)h1->pps_buffers, + MAX_PPS_COUNT, sizeof(PPS)); + h->pps = h1->pps; - //Dequantization matrices - //FIXME these are big - can they be only copied when PPS changes? + // Dequantization matrices + // FIXME these are big - can they be only copied when PPS changes? copy_fields(h, h1, dequant4_buffer, dequant4_coeff); - for(i=0; i<6; i++) - h->dequant4_coeff[i] = h->dequant4_buffer[0] + (h1->dequant4_coeff[i] - h1->dequant4_buffer[0]); + for (i = 0; i < 6; i++) + h->dequant4_coeff[i] = h->dequant4_buffer[0] + + (h1->dequant4_coeff[i] - h1->dequant4_buffer[0]); - for(i=0; i<6; i++) - h->dequant8_coeff[i] = h->dequant8_buffer[0] + (h1->dequant8_coeff[i] - h1->dequant8_buffer[0]); + for (i = 0; i < 6; i++) + h->dequant8_coeff[i] = h->dequant8_buffer[0] + + (h1->dequant8_coeff[i] - h1->dequant8_buffer[0]); h->dequant_coeff_pps = h1->dequant_coeff_pps; - //POC timing + // POC timing copy_fields(h, h1, poc_lsb, redundant_pic_count); - //reference lists + // reference lists copy_fields(h, h1, ref_count, list_count); - copy_fields(h, h1, ref_list, intra_gb); + copy_fields(h, h1, ref_list, intra_gb); copy_fields(h, h1, short_ref, cabac_init_idc); - copy_picture_range(h->short_ref, h1->short_ref, 32, s, s1); - copy_picture_range(h->long_ref, h1->long_ref, 32, s, s1); - copy_picture_range(h->delayed_pic, h1->delayed_pic, MAX_DELAYED_PIC_COUNT+2, s, s1); + copy_picture_range(h->short_ref, h1->short_ref, 32, s, s1); + copy_picture_range(h->long_ref, h1->long_ref, 32, s, s1); + copy_picture_range(h->delayed_pic, h1->delayed_pic, + MAX_DELAYED_PIC_COUNT + 2, s, s1); h->last_slice_type = h1->last_slice_type; - if(!s->current_picture_ptr) return 0; + if (!s->current_picture_ptr) + return 0; - if(!s->dropable) { + if (!s->dropable) { err = ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index); - h->prev_poc_msb = h->poc_msb; - h->prev_poc_lsb = h->poc_lsb; + h->prev_poc_msb = h->poc_msb; + h->prev_poc_lsb = h->poc_lsb; } - h->prev_frame_num_offset= h->frame_num_offset; - h->prev_frame_num = h->frame_num; - h->outputed_poc = h->next_outputed_poc; + h->prev_frame_num_offset = h->frame_num_offset; + h->prev_frame_num = h->frame_num; + h->outputed_poc = h->next_outputed_poc; return err; } -int ff_h264_frame_start(H264Context *h){ - MpegEncContext * const s = &h->s; +int ff_h264_frame_start(H264Context *h) +{ + MpegEncContext *const s = &h->s; int i; const int pixel_shift = h->pixel_shift; - if(ff_MPV_frame_start(s, s->avctx) < 0) + if (ff_MPV_frame_start(s, s->avctx) < 0) return -1; ff_er_frame_start(s); /* @@ -1291,61 +1412,65 @@ int ff_h264_frame_start(H264Context *h){ * See decode_nal_units(). */ s->current_picture_ptr->f.key_frame = 0; - s->current_picture_ptr->mmco_reset= 0; + s->current_picture_ptr->mmco_reset = 0; assert(s->linesize && s->uvlinesize); - for(i=0; i<16; i++){ - h->block_offset[i]= (4*((scan8[i] - scan8[0])&7) << pixel_shift) + 4*s->linesize*((scan8[i] - scan8[0])>>3); - h->block_offset[48+i]= (4*((scan8[i] - scan8[0])&7) << pixel_shift) + 8*s->linesize*((scan8[i] - scan8[0])>>3); + for (i = 0; i < 16; i++) { + h->block_offset[i] = (4 * ((scan8[i] - scan8[0]) & 7) << pixel_shift) + 4 * s->linesize * ((scan8[i] - scan8[0]) >> 3); + h->block_offset[48 + i] = (4 * ((scan8[i] - scan8[0]) & 7) << pixel_shift) + 8 * s->linesize * ((scan8[i] - scan8[0]) >> 3); } - for(i=0; i<16; i++){ - h->block_offset[16+i]= - h->block_offset[32+i]= (4*((scan8[i] - scan8[0])&7) << pixel_shift) + 4*s->uvlinesize*((scan8[i] - scan8[0])>>3); - h->block_offset[48+16+i]= - h->block_offset[48+32+i]= (4*((scan8[i] - scan8[0])&7) << pixel_shift) + 8*s->uvlinesize*((scan8[i] - scan8[0])>>3); + for (i = 0; i < 16; i++) { + h->block_offset[16 + i] = + h->block_offset[32 + i] = (4 * ((scan8[i] - scan8[0]) & 7) << pixel_shift) + 4 * s->uvlinesize * ((scan8[i] - scan8[0]) >> 3); + h->block_offset[48 + 16 + i] = + h->block_offset[48 + 32 + i] = (4 * ((scan8[i] - scan8[0]) & 7) << pixel_shift) + 8 * s->uvlinesize * ((scan8[i] - scan8[0]) >> 3); } /* can't be in alloc_tables because linesize isn't known there. * FIXME: redo bipred weight to not require extra buffer? */ - for(i = 0; i < s->slice_context_count; i++) - if(h->thread_context[i] && !h->thread_context[i]->s.obmc_scratchpad) - h->thread_context[i]->s.obmc_scratchpad = av_malloc(16*6*s->linesize); + for (i = 0; i < s->slice_context_count; i++) + if (h->thread_context[i] && !h->thread_context[i]->s.obmc_scratchpad) + h->thread_context[i]->s.obmc_scratchpad = av_malloc(16 * 6 * s->linesize); - /* some macroblocks can be accessed before they're available in case of lost slices, mbaff or threading*/ - memset(h->slice_table, -1, (s->mb_height*s->mb_stride-1) * sizeof(*h->slice_table)); + /* Some macroblocks can be accessed before they're available in case + * of lost slices, MBAFF or threading. */ + memset(h->slice_table, -1, + (s->mb_height * s->mb_stride - 1) * sizeof(*h->slice_table)); -// s->decode = (s->flags & CODEC_FLAG_PSNR) || !s->encoding || s->current_picture.f.reference /*|| h->contains_intra*/ || 1; + // s->decode = (s->flags & CODEC_FLAG_PSNR) || !s->encoding || + // s->current_picture.f.reference /* || h->contains_intra */ || 1; - // We mark the current picture as non-reference after allocating it, so - // that if we break out due to an error it can be released automatically - // in the next ff_MPV_frame_start(). - // SVQ3 as well as most other codecs have only last/next/current and thus - // get released even with set reference, besides SVQ3 and others do not - // mark frames as reference later "naturally". - if(s->codec_id != CODEC_ID_SVQ3) + /* We mark the current picture as non-reference after allocating it, so + * that if we break out due to an error it can be released automatically + * in the next ff_MPV_frame_start(). + * SVQ3 as well as most other codecs have only last/next/current and thus + * get released even with set reference, besides SVQ3 and others do not + * mark frames as reference later "naturally". */ + if (s->codec_id != CODEC_ID_SVQ3) s->current_picture_ptr->f.reference = 0; - s->current_picture_ptr->field_poc[0]= - s->current_picture_ptr->field_poc[1]= INT_MAX; + s->current_picture_ptr->field_poc[0] = + s->current_picture_ptr->field_poc[1] = INT_MAX; h->next_output_pic = NULL; - assert(s->current_picture_ptr->long_ref==0); + assert(s->current_picture_ptr->long_ref == 0); return 0; } /** - * Run setup operations that must be run after slice header decoding. - * This includes finding the next displayed frame. - * - * @param h h264 master context - * @param setup_finished enough NALs have been read that we can call - * ff_thread_finish_setup() - */ -static void decode_postinit(H264Context *h, int setup_finished){ - MpegEncContext * const s = &h->s; + * Run setup operations that must be run after slice header decoding. + * This includes finding the next displayed frame. + * + * @param h h264 master context + * @param setup_finished enough NALs have been read that we can call + * ff_thread_finish_setup() + */ +static void decode_postinit(H264Context *h, int setup_finished) +{ + MpegEncContext *const s = &h->s; Picture *out = s->current_picture_ptr; Picture *cur = s->current_picture_ptr; int i, pics, out_of_order, out_idx; @@ -1354,13 +1479,15 @@ static void decode_postinit(H264Context *h, int setup_finished){ s->current_picture_ptr->f.qscale_type = FF_QSCALE_TYPE_H264; s->current_picture_ptr->f.pict_type = s->pict_type; - if (h->next_output_pic) return; + if (h->next_output_pic) + return; - if (cur->field_poc[0]==INT_MAX || cur->field_poc[1]==INT_MAX) { - //FIXME: if we have two PAFF fields in one packet, we can't start the next thread here. - //If we have one field per packet, we can. The check in decode_nal_units() is not good enough - //to find this yet, so we assume the worst for now. - //if (setup_finished) + if (cur->field_poc[0] == INT_MAX || cur->field_poc[1] == INT_MAX) { + /* FIXME: if we have two PAFF fields in one packet, we can't start + * the next thread here. If we have one field per packet, we can. + * The check in decode_nal_units() is not good enough to find this + * yet, so we assume the worst for now. */ + // if (setup_finished) // ff_thread_finish_setup(s->avctx); return; } @@ -1369,11 +1496,11 @@ static void decode_postinit(H264Context *h, int setup_finished){ cur->f.repeat_pict = 0; /* Signal interlacing information externally. */ - /* Prioritize picture timing SEI information over used decoding process if it exists. */ + /* Prioritize picture timing SEI information over used + * decoding process if it exists. */ - if(h->sps.pic_struct_present_flag){ - switch (h->sei_pic_struct) - { + if (h->sps.pic_struct_present_flag) { + switch (h->sei_pic_struct) { case SEI_PIC_STRUCT_FRAME: break; case SEI_PIC_STRUCT_TOP_FIELD: @@ -1390,12 +1517,13 @@ static void decode_postinit(H264Context *h, int setup_finished){ break; case SEI_PIC_STRUCT_TOP_BOTTOM_TOP: case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM: - // Signal the possibility of telecined film externally (pic_struct 5,6) - // From these hints, let the applications decide if they apply deinterlacing. + /* Signal the possibility of telecined film externally + * (pic_struct 5,6). From these hints, let the applications + * decide if they apply deinterlacing. */ cur->f.repeat_pict = 1; break; case SEI_PIC_STRUCT_FRAME_DOUBLING: - // Force progressive here, as doubling interlaced frame is a bad idea. + // Force progressive here, doubling interlaced frame is a bad idea. cur->f.repeat_pict = 2; break; case SEI_PIC_STRUCT_FRAME_TRIPLING: @@ -1403,49 +1531,52 @@ static void decode_postinit(H264Context *h, int setup_finished){ break; } - if ((h->sei_ct_type & 3) && h->sei_pic_struct <= SEI_PIC_STRUCT_BOTTOM_TOP) + if ((h->sei_ct_type & 3) && + h->sei_pic_struct <= SEI_PIC_STRUCT_BOTTOM_TOP) cur->f.interlaced_frame = (h->sei_ct_type & (1 << 1)) != 0; - }else{ + } else { /* Derive interlacing flag from used decoding process. */ cur->f.interlaced_frame = FIELD_OR_MBAFF_PICTURE; } h->prev_interlaced_frame = cur->f.interlaced_frame; - if (cur->field_poc[0] != cur->field_poc[1]){ + if (cur->field_poc[0] != cur->field_poc[1]) { /* Derive top_field_first from field pocs. */ cur->f.top_field_first = cur->field_poc[0] < cur->field_poc[1]; - }else{ + } else { if (cur->f.interlaced_frame || h->sps.pic_struct_present_flag) { - /* Use picture timing SEI information. Even if it is a information of a past frame, better than nothing. */ - if(h->sei_pic_struct == SEI_PIC_STRUCT_TOP_BOTTOM - || h->sei_pic_struct == SEI_PIC_STRUCT_TOP_BOTTOM_TOP) + /* Use picture timing SEI information. Even if it is a + * information of a past frame, better than nothing. */ + if (h->sei_pic_struct == SEI_PIC_STRUCT_TOP_BOTTOM || + h->sei_pic_struct == SEI_PIC_STRUCT_TOP_BOTTOM_TOP) cur->f.top_field_first = 1; else cur->f.top_field_first = 0; - }else{ + } else { /* Most likely progressive */ cur->f.top_field_first = 0; } } - //FIXME do something with unavailable reference frames + // FIXME do something with unavailable reference frames /* Sort B-frames into display order */ - if(h->sps.bitstream_restriction_flag - && s->avctx->has_b_frames < h->sps.num_reorder_frames){ + if (h->sps.bitstream_restriction_flag && + s->avctx->has_b_frames < h->sps.num_reorder_frames) { s->avctx->has_b_frames = h->sps.num_reorder_frames; - s->low_delay = 0; + s->low_delay = 0; } - if( s->avctx->strict_std_compliance >= FF_COMPLIANCE_STRICT - && !h->sps.bitstream_restriction_flag){ + if (s->avctx->strict_std_compliance >= FF_COMPLIANCE_STRICT && + !h->sps.bitstream_restriction_flag) { s->avctx->has_b_frames = MAX_DELAYED_PIC_COUNT - 1; - s->low_delay= 0; + s->low_delay = 0; } pics = 0; - while(h->delayed_pic[pics]) pics++; + while (h->delayed_pic[pics]) + pics++; assert(pics <= MAX_DELAYED_PIC_COUNT); @@ -1470,7 +1601,8 @@ static void decode_postinit(H264Context *h, int setup_finished){ cnt += out->poc < h->last_pocs[i]; invalid += out->poc == INT_MIN; } - if (!h->mmco_reset && !cur->f.key_frame && cnt + invalid == MAX_DELAYED_PIC_COUNT && cnt > 0) { + if (!h->mmco_reset && !cur->f.key_frame && + cnt + invalid == MAX_DELAYED_PIC_COUNT && cnt > 0) { h->mmco_reset = 2; if (pics > 1) h->delayed_pic[pics - 2]->mmco_reset = 2; @@ -1481,49 +1613,55 @@ static void decode_postinit(H264Context *h, int setup_finished){ cnt = 0; invalid = MAX_DELAYED_PIC_COUNT; } - out = h->delayed_pic[0]; + out = h->delayed_pic[0]; out_idx = 0; - for (i = 1; i < MAX_DELAYED_PIC_COUNT && h->delayed_pic[i] && - !h->delayed_pic[i-1]->mmco_reset && !h->delayed_pic[i]->f.key_frame; i++) - { - if(h->delayed_pic[i]->poc < out->poc){ - out = h->delayed_pic[i]; + for (i = 1; i < MAX_DELAYED_PIC_COUNT && + h->delayed_pic[i] && + !h->delayed_pic[i - 1]->mmco_reset && + !h->delayed_pic[i]->f.key_frame; + i++) + if (h->delayed_pic[i]->poc < out->poc) { + out = h->delayed_pic[i]; out_idx = i; } - } - if (s->avctx->has_b_frames == 0 && (h->delayed_pic[0]->f.key_frame || h->mmco_reset)) + if (s->avctx->has_b_frames == 0 && + (h->delayed_pic[0]->f.key_frame || h->mmco_reset)) h->next_outputed_poc = INT_MIN; - out_of_order = !out->f.key_frame && !h->mmco_reset && (out->poc < h->next_outputed_poc); + out_of_order = !out->f.key_frame && !h->mmco_reset && + (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) { + 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) { if (invalid + cnt < MAX_DELAYED_PIC_COUNT) { s->avctx->has_b_frames = FFMAX(s->avctx->has_b_frames, cnt); } s->low_delay = 0; } else if (s->low_delay && - ((h->next_outputed_poc != INT_MIN && out->poc > h->next_outputed_poc + 2) || + ((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(pics > s->avctx->has_b_frames){ + if (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 - // or else the first thread can release the picture and reuse it unsafely - for(i=out_idx; h->delayed_pic[i]; i++) - h->delayed_pic[i] = h->delayed_pic[i+1]; + // for frame threading, the owner must be the second field's thread or + // else the first thread can release the picture and reuse it unsafely + out->owner2 = s; + 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)); + 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] = cur->poc; - if(!out_of_order && pics > s->avctx->has_b_frames){ + if (!out_of_order && pics > s->avctx->has_b_frames) { h->next_output_pic = out; if (out->mmco_reset) { if (out_idx > 0) { - h->next_outputed_poc = out->poc; + h->next_outputed_poc = out->poc; h->delayed_pic[out_idx - 1]->mmco_reset = out->mmco_reset; } else { h->next_outputed_poc = INT_MIN; @@ -1536,7 +1674,7 @@ static void decode_postinit(H264Context *h, int setup_finished){ } } h->mmco_reset = 0; - }else{ + } else { av_log(s->avctx, AV_LOG_DEBUG, "no picture\n"); } @@ -1546,122 +1684,124 @@ static void decode_postinit(H264Context *h, int setup_finished){ static av_always_inline void backup_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, - int linesize, int uvlinesize, int simple) + int linesize, int uvlinesize, + int simple) { - MpegEncContext * const s = &h->s; + MpegEncContext *const s = &h->s; uint8_t *top_border; int top_idx = 1; const int pixel_shift = h->pixel_shift; int chroma444 = CHROMA444; int chroma422 = CHROMA422; - src_y -= linesize; + src_y -= linesize; src_cb -= uvlinesize; src_cr -= uvlinesize; - if(!simple && FRAME_MBAFF){ - if(s->mb_y&1){ - if(!MB_MBAFF){ + if (!simple && FRAME_MBAFF) { + if (s->mb_y & 1) { + if (!MB_MBAFF) { top_border = h->top_borders[0][s->mb_x]; - AV_COPY128(top_border, src_y + 15*linesize); + AV_COPY128(top_border, src_y + 15 * linesize); if (pixel_shift) - AV_COPY128(top_border+16, src_y+15*linesize+16); - if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){ - if(chroma444){ - if (pixel_shift){ - AV_COPY128(top_border+32, src_cb + 15*uvlinesize); - AV_COPY128(top_border+48, src_cb + 15*uvlinesize+16); - AV_COPY128(top_border+64, src_cr + 15*uvlinesize); - AV_COPY128(top_border+80, src_cr + 15*uvlinesize+16); - } else { - AV_COPY128(top_border+16, src_cb + 15*uvlinesize); - AV_COPY128(top_border+32, src_cr + 15*uvlinesize); - } - } else if(chroma422) { + AV_COPY128(top_border + 16, src_y + 15 * linesize + 16); + if (simple || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) { + if (chroma444) { if (pixel_shift) { - AV_COPY128(top_border+32, src_cb + 15*uvlinesize); - AV_COPY128(top_border+48, src_cr + 15*uvlinesize); + AV_COPY128(top_border + 32, src_cb + 15 * uvlinesize); + AV_COPY128(top_border + 48, src_cb + 15 * uvlinesize + 16); + AV_COPY128(top_border + 64, src_cr + 15 * uvlinesize); + AV_COPY128(top_border + 80, src_cr + 15 * uvlinesize + 16); } else { - AV_COPY64(top_border+16, src_cb + 15*uvlinesize); - AV_COPY64(top_border+24, src_cr + 15*uvlinesize); + AV_COPY128(top_border + 16, src_cb + 15 * uvlinesize); + AV_COPY128(top_border + 32, src_cr + 15 * uvlinesize); + } + } else if (chroma422) { + if (pixel_shift) { + AV_COPY128(top_border + 32, src_cb + 15 * uvlinesize); + AV_COPY128(top_border + 48, src_cr + 15 * uvlinesize); + } else { + AV_COPY64(top_border + 16, src_cb + 15 * uvlinesize); + AV_COPY64(top_border + 24, src_cr + 15 * uvlinesize); } } else { if (pixel_shift) { - AV_COPY128(top_border+32, src_cb+7*uvlinesize); - AV_COPY128(top_border+48, src_cr+7*uvlinesize); + AV_COPY128(top_border + 32, src_cb + 7 * uvlinesize); + AV_COPY128(top_border + 48, src_cr + 7 * uvlinesize); } else { - AV_COPY64(top_border+16, src_cb+7*uvlinesize); - AV_COPY64(top_border+24, src_cr+7*uvlinesize); + AV_COPY64(top_border + 16, src_cb + 7 * uvlinesize); + AV_COPY64(top_border + 24, src_cr + 7 * uvlinesize); } } } } - }else if(MB_MBAFF){ + } else if (MB_MBAFF) { top_idx = 0; - }else + } else return; } top_border = h->top_borders[top_idx][s->mb_x]; - // There are two lines saved, the line above the the top macroblock of a pair, - // and the line above the bottom macroblock - AV_COPY128(top_border, src_y + 16*linesize); + /* There are two lines saved, the line above the the top macroblock + * of a pair, and the line above the bottom macroblock. */ + AV_COPY128(top_border, src_y + 16 * linesize); if (pixel_shift) - AV_COPY128(top_border+16, src_y+16*linesize+16); + AV_COPY128(top_border + 16, src_y + 16 * linesize + 16); - if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){ - if(chroma444){ - if (pixel_shift){ - AV_COPY128(top_border+32, src_cb + 16*linesize); - AV_COPY128(top_border+48, src_cb + 16*linesize+16); - AV_COPY128(top_border+64, src_cr + 16*linesize); - AV_COPY128(top_border+80, src_cr + 16*linesize+16); - } else { - AV_COPY128(top_border+16, src_cb + 16*linesize); - AV_COPY128(top_border+32, src_cr + 16*linesize); - } - } else if(chroma422) { + if (simple || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) { + if (chroma444) { if (pixel_shift) { - AV_COPY128(top_border+32, src_cb+16*uvlinesize); - AV_COPY128(top_border+48, src_cr+16*uvlinesize); + AV_COPY128(top_border + 32, src_cb + 16 * linesize); + AV_COPY128(top_border + 48, src_cb + 16 * linesize + 16); + AV_COPY128(top_border + 64, src_cr + 16 * linesize); + AV_COPY128(top_border + 80, src_cr + 16 * linesize + 16); } else { - AV_COPY64(top_border+16, src_cb+16*uvlinesize); - AV_COPY64(top_border+24, src_cr+16*uvlinesize); + AV_COPY128(top_border + 16, src_cb + 16 * linesize); + AV_COPY128(top_border + 32, src_cr + 16 * linesize); + } + } else if (chroma422) { + if (pixel_shift) { + AV_COPY128(top_border + 32, src_cb + 16 * uvlinesize); + AV_COPY128(top_border + 48, src_cr + 16 * uvlinesize); + } else { + AV_COPY64(top_border + 16, src_cb + 16 * uvlinesize); + AV_COPY64(top_border + 24, src_cr + 16 * uvlinesize); } } else { if (pixel_shift) { - AV_COPY128(top_border+32, src_cb+8*uvlinesize); - AV_COPY128(top_border+48, src_cr+8*uvlinesize); + AV_COPY128(top_border + 32, src_cb + 8 * uvlinesize); + AV_COPY128(top_border + 48, src_cr + 8 * uvlinesize); } else { - AV_COPY64(top_border+16, src_cb+8*uvlinesize); - AV_COPY64(top_border+24, src_cr+8*uvlinesize); + AV_COPY64(top_border + 16, src_cb + 8 * uvlinesize); + AV_COPY64(top_border + 24, src_cr + 8 * uvlinesize); } } } } static av_always_inline void xchg_mb_border(H264Context *h, uint8_t *src_y, - uint8_t *src_cb, uint8_t *src_cr, - int linesize, int uvlinesize, - int xchg, int chroma444, - int simple, int pixel_shift){ - MpegEncContext * const s = &h->s; + uint8_t *src_cb, uint8_t *src_cr, + int linesize, int uvlinesize, + int xchg, int chroma444, + int simple, int pixel_shift) +{ + MpegEncContext *const s = &h->s; int deblock_topleft; int deblock_top; int top_idx = 1; uint8_t *top_border_m1; uint8_t *top_border; - if(!simple && FRAME_MBAFF){ - if(s->mb_y&1){ - if(!MB_MBAFF) + if (!simple && FRAME_MBAFF) { + if (s->mb_y & 1) { + if (!MB_MBAFF) return; - }else{ + } else { top_idx = MB_MBAFF ? 0 : 1; } } - if(h->deblocking_filter == 2) { + if (h->deblocking_filter == 2) { deblock_topleft = h->slice_table[h->mb_xy - 1 - s->mb_stride] == h->slice_num; deblock_top = h->top_type; } else { @@ -1669,38 +1809,41 @@ static av_always_inline void xchg_mb_border(H264Context *h, uint8_t *src_y, deblock_top = (s->mb_y > !!MB_FIELD); } - src_y -= linesize + 1 + pixel_shift; + src_y -= linesize + 1 + pixel_shift; src_cb -= uvlinesize + 1 + pixel_shift; src_cr -= uvlinesize + 1 + pixel_shift; - top_border_m1 = h->top_borders[top_idx][s->mb_x-1]; + top_border_m1 = h->top_borders[top_idx][s->mb_x - 1]; top_border = h->top_borders[top_idx][s->mb_x]; -#define XCHG(a,b,xchg)\ - if (pixel_shift) {\ - if (xchg) {\ - AV_SWAP64(b+0,a+0);\ - AV_SWAP64(b+8,a+8);\ - } else {\ - AV_COPY128(b,a); \ - }\ - } else \ -if (xchg) AV_SWAP64(b,a);\ -else AV_COPY64(b,a); +#define XCHG(a, b, xchg) \ + if (pixel_shift) { \ + if (xchg) { \ + AV_SWAP64(b + 0, a + 0); \ + AV_SWAP64(b + 8, a + 8); \ + } else { \ + AV_COPY128(b, a); \ + } \ + } else if (xchg) \ + AV_SWAP64(b, a); \ + else \ + AV_COPY64(b, a); - if(deblock_top){ - if(deblock_topleft){ - XCHG(top_border_m1 + (8 << pixel_shift), src_y - (7 << pixel_shift), 1); + if (deblock_top) { + if (deblock_topleft) { + XCHG(top_border_m1 + (8 << pixel_shift), + src_y - (7 << pixel_shift), 1); } XCHG(top_border + (0 << pixel_shift), src_y + (1 << pixel_shift), xchg); XCHG(top_border + (8 << pixel_shift), src_y + (9 << pixel_shift), 1); - if(s->mb_x+1 < s->mb_width){ - XCHG(h->top_borders[top_idx][s->mb_x+1], src_y + (17 << pixel_shift), 1); + if (s->mb_x + 1 < s->mb_width) { + XCHG(h->top_borders[top_idx][s->mb_x + 1], + src_y + (17 << pixel_shift), 1); } } - if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){ - if(chroma444){ - if(deblock_topleft){ + if (simple || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) { + if (chroma444) { + if (deblock_topleft) { XCHG(top_border_m1 + (24 << pixel_shift), src_cb - (7 << pixel_shift), 1); XCHG(top_border_m1 + (40 << pixel_shift), src_cr - (7 << pixel_shift), 1); } @@ -1708,198 +1851,239 @@ else AV_COPY64(b,a); XCHG(top_border + (24 << pixel_shift), src_cb + (9 << pixel_shift), 1); XCHG(top_border + (32 << pixel_shift), src_cr + (1 << pixel_shift), xchg); XCHG(top_border + (40 << pixel_shift), src_cr + (9 << pixel_shift), 1); - if(s->mb_x+1 < s->mb_width){ - XCHG(h->top_borders[top_idx][s->mb_x+1] + (16 << pixel_shift), src_cb + (17 << pixel_shift), 1); - XCHG(h->top_borders[top_idx][s->mb_x+1] + (32 << pixel_shift), src_cr + (17 << pixel_shift), 1); + if (s->mb_x + 1 < s->mb_width) { + XCHG(h->top_borders[top_idx][s->mb_x + 1] + (16 << pixel_shift), src_cb + (17 << pixel_shift), 1); + XCHG(h->top_borders[top_idx][s->mb_x + 1] + (32 << pixel_shift), src_cr + (17 << pixel_shift), 1); } } else { - if(deblock_top){ - if(deblock_topleft){ + if (deblock_top) { + if (deblock_topleft) { XCHG(top_border_m1 + (16 << pixel_shift), src_cb - (7 << pixel_shift), 1); XCHG(top_border_m1 + (24 << pixel_shift), src_cr - (7 << pixel_shift), 1); } - XCHG(top_border + (16 << pixel_shift), src_cb+1+pixel_shift, 1); - XCHG(top_border + (24 << pixel_shift), src_cr+1+pixel_shift, 1); + XCHG(top_border + (16 << pixel_shift), src_cb + 1 + pixel_shift, 1); + XCHG(top_border + (24 << pixel_shift), src_cr + 1 + pixel_shift, 1); } } } } -static av_always_inline int dctcoef_get(DCTELEM *mb, int high_bit_depth, int index) { +static av_always_inline int dctcoef_get(DCTELEM *mb, int high_bit_depth, + int index) +{ if (high_bit_depth) { - return AV_RN32A(((int32_t*)mb) + index); + return AV_RN32A(((int32_t *)mb) + index); } else return AV_RN16A(mb + index); } -static av_always_inline void dctcoef_set(DCTELEM *mb, int high_bit_depth, int index, int value) { +static av_always_inline void dctcoef_set(DCTELEM *mb, int high_bit_depth, + int index, int value) +{ if (high_bit_depth) { - AV_WN32A(((int32_t*)mb) + index, value); + AV_WN32A(((int32_t *)mb) + index, value); } else AV_WN16A(mb + index, value); } -static av_always_inline void hl_decode_mb_predict_luma(H264Context *h, int mb_type, int is_h264, int simple, int transform_bypass, - int pixel_shift, int *block_offset, int linesize, uint8_t *dest_y, int p) +static av_always_inline void hl_decode_mb_predict_luma(H264Context *h, + int mb_type, int is_h264, + int simple, + int transform_bypass, + int pixel_shift, + int *block_offset, + int linesize, + uint8_t *dest_y, int p) { - MpegEncContext * const s = &h->s; + MpegEncContext *const s = &h->s; void (*idct_add)(uint8_t *dst, DCTELEM *block, int stride); void (*idct_dc_add)(uint8_t *dst, DCTELEM *block, int stride); int i; - int qscale = p == 0 ? s->qscale : h->chroma_qp[p-1]; - block_offset += 16*p; - if(IS_INTRA4x4(mb_type)){ - if(simple || !s->encoding){ - if(IS_8x8DCT(mb_type)){ - if(transform_bypass){ - idct_dc_add = - idct_add = s->dsp.add_pixels8; - }else{ + int qscale = p == 0 ? s->qscale : h->chroma_qp[p - 1]; + block_offset += 16 * p; + if (IS_INTRA4x4(mb_type)) { + if (simple || !s->encoding) { + if (IS_8x8DCT(mb_type)) { + if (transform_bypass) { + idct_dc_add = + idct_add = s->dsp.add_pixels8; + } else { idct_dc_add = h->h264dsp.h264_idct8_dc_add; idct_add = h->h264dsp.h264_idct8_add; } - for(i=0; i<16; i+=4){ - uint8_t * const ptr= dest_y + block_offset[i]; - const int dir= h->intra4x4_pred_mode_cache[ scan8[i] ]; - if(transform_bypass && h->sps.profile_idc==244 && dir<=1){ - h->hpc.pred8x8l_add[dir](ptr, h->mb + (i*16+p*256 << pixel_shift), linesize); - }else{ - const int nnz = h->non_zero_count_cache[ scan8[i+p*16] ]; - h->hpc.pred8x8l[ dir ](ptr, (h->topleft_samples_available<topright_samples_available<mb, pixel_shift, i*16+p*256)) - idct_dc_add(ptr, h->mb + (i*16+p*256 << pixel_shift), linesize); + for (i = 0; i < 16; i += 4) { + uint8_t *const ptr = dest_y + block_offset[i]; + const int dir = h->intra4x4_pred_mode_cache[scan8[i]]; + if (transform_bypass && h->sps.profile_idc == 244 && dir <= 1) { + h->hpc.pred8x8l_add[dir](ptr, h->mb + (i * 16 + p * 256 << pixel_shift), linesize); + } else { + const int nnz = h->non_zero_count_cache[scan8[i + p * 16]]; + h->hpc.pred8x8l[dir](ptr, (h->topleft_samples_available << i) & 0x8000, + (h->topright_samples_available << i) & 0x4000, linesize); + if (nnz) { + if (nnz == 1 && dctcoef_get(h->mb, pixel_shift, i * 16 + p * 256)) + idct_dc_add(ptr, h->mb + (i * 16 + p * 256 << pixel_shift), linesize); else - idct_add (ptr, h->mb + (i*16+p*256 << pixel_shift), linesize); + idct_add(ptr, h->mb + (i * 16 + p * 256 << pixel_shift), linesize); } } } - }else{ - if(transform_bypass){ - idct_dc_add = - idct_add = s->dsp.add_pixels4; - }else{ + } else { + if (transform_bypass) { + idct_dc_add = + idct_add = s->dsp.add_pixels4; + } else { idct_dc_add = h->h264dsp.h264_idct_dc_add; idct_add = h->h264dsp.h264_idct_add; } - for(i=0; i<16; i++){ - uint8_t * const ptr= dest_y + block_offset[i]; - const int dir= h->intra4x4_pred_mode_cache[ scan8[i] ]; + for (i = 0; i < 16; i++) { + uint8_t *const ptr = dest_y + block_offset[i]; + const int dir = h->intra4x4_pred_mode_cache[scan8[i]]; - if(transform_bypass && h->sps.profile_idc==244 && dir<=1){ - h->hpc.pred4x4_add[dir](ptr, h->mb + (i*16+p*256 << pixel_shift), linesize); - }else{ + if (transform_bypass && h->sps.profile_idc == 244 && dir <= 1) { + h->hpc.pred4x4_add[dir](ptr, h->mb + (i * 16 + p * 256 << pixel_shift), linesize); + } else { uint8_t *topright; int nnz, tr; uint64_t tr_high; - if(dir == DIAG_DOWN_LEFT_PRED || dir == VERT_LEFT_PRED){ - const int topright_avail= (h->topright_samples_available<topright_samples_available << i) & 0x8000; assert(s->mb_y || linesize <= block_offset[i]); - if(!topright_avail){ + if (!topright_avail) { if (pixel_shift) { - tr_high= ((uint16_t*)ptr)[3 - linesize/2]*0x0001000100010001ULL; - topright= (uint8_t*) &tr_high; + tr_high = ((uint16_t *)ptr)[3 - linesize / 2] * 0x0001000100010001ULL; + topright = (uint8_t *)&tr_high; } else { - tr= ptr[3 - linesize]*0x01010101u; - topright= (uint8_t*) &tr; + tr = ptr[3 - linesize] * 0x01010101u; + topright = (uint8_t *)&tr; } - }else - topright= ptr + (4 << pixel_shift) - linesize; - }else - topright= NULL; + } else + topright = ptr + (4 << pixel_shift) - linesize; + } else + topright = NULL; - h->hpc.pred4x4[ dir ](ptr, topright, linesize); - nnz = h->non_zero_count_cache[ scan8[i+p*16] ]; - if(nnz){ - if(is_h264){ - if(nnz == 1 && dctcoef_get(h->mb, pixel_shift, i*16+p*256)) - idct_dc_add(ptr, h->mb + (i*16+p*256 << pixel_shift), linesize); + h->hpc.pred4x4[dir](ptr, topright, linesize); + nnz = h->non_zero_count_cache[scan8[i + p * 16]]; + if (nnz) { + if (is_h264) { + if (nnz == 1 && dctcoef_get(h->mb, pixel_shift, i * 16 + p * 256)) + idct_dc_add(ptr, h->mb + (i * 16 + p * 256 << pixel_shift), linesize); else - idct_add (ptr, h->mb + (i*16+p*256 << pixel_shift), linesize); + idct_add(ptr, h->mb + (i * 16 + p * 256 << pixel_shift), linesize); } else if (CONFIG_SVQ3_DECODER) - ff_svq3_add_idct_c(ptr, h->mb + i*16+p*256, linesize, qscale, 0); + ff_svq3_add_idct_c(ptr, h->mb + i * 16 + p * 256, linesize, qscale, 0); } } } } } - }else{ - h->hpc.pred16x16[ h->intra16x16_pred_mode ](dest_y , linesize); - if(is_h264){ - if(h->non_zero_count_cache[ scan8[LUMA_DC_BLOCK_INDEX+p] ]){ - if(!transform_bypass) - h->h264dsp.h264_luma_dc_dequant_idct(h->mb+(p*256 << pixel_shift), h->mb_luma_dc[p], h->dequant4_coeff[p][qscale][0]); - else{ - static const uint8_t dc_mapping[16] = { 0*16, 1*16, 4*16, 5*16, 2*16, 3*16, 6*16, 7*16, - 8*16, 9*16,12*16,13*16,10*16,11*16,14*16,15*16}; - for(i = 0; i < 16; i++) - dctcoef_set(h->mb+(p*256 << pixel_shift), pixel_shift, dc_mapping[i], dctcoef_get(h->mb_luma_dc[p], pixel_shift, i)); + } else { + h->hpc.pred16x16[h->intra16x16_pred_mode](dest_y, linesize); + if (is_h264) { + if (h->non_zero_count_cache[scan8[LUMA_DC_BLOCK_INDEX + p]]) { + if (!transform_bypass) + h->h264dsp.h264_luma_dc_dequant_idct(h->mb + (p * 256 << pixel_shift), + h->mb_luma_dc[p], + h->dequant4_coeff[p][qscale][0]); + else { + static const uint8_t dc_mapping[16] = { + 0 * 16, 1 * 16, 4 * 16, 5 * 16, + 2 * 16, 3 * 16, 6 * 16, 7 * 16, + 8 * 16, 9 * 16, 12 * 16, 13 * 16, + 10 * 16, 11 * 16, 14 * 16, 15 * 16 }; + for (i = 0; i < 16; i++) + dctcoef_set(h->mb + (p * 256 << pixel_shift), + pixel_shift, dc_mapping[i], + dctcoef_get(h->mb_luma_dc[p], + pixel_shift, i)); } } } else if (CONFIG_SVQ3_DECODER) - ff_svq3_luma_dc_dequant_idct_c(h->mb+p*256, h->mb_luma_dc[p], qscale); + ff_svq3_luma_dc_dequant_idct_c(h->mb + p * 256, + h->mb_luma_dc[p], qscale); } } -static av_always_inline void hl_decode_mb_idct_luma(H264Context *h, int mb_type, int is_h264, int simple, int transform_bypass, - int pixel_shift, int *block_offset, int linesize, uint8_t *dest_y, int p) +static av_always_inline void hl_decode_mb_idct_luma(H264Context *h, int mb_type, + int is_h264, int simple, + int transform_bypass, + int pixel_shift, + int *block_offset, + int linesize, + uint8_t *dest_y, int p) { - MpegEncContext * const s = &h->s; + MpegEncContext *const s = &h->s; void (*idct_add)(uint8_t *dst, DCTELEM *block, int stride); int i; - block_offset += 16*p; - if(!IS_INTRA4x4(mb_type)){ - if(is_h264){ - if(IS_INTRA16x16(mb_type)){ - if(transform_bypass){ - if(h->sps.profile_idc==244 && (h->intra16x16_pred_mode==VERT_PRED8x8 || h->intra16x16_pred_mode==HOR_PRED8x8)){ - h->hpc.pred16x16_add[h->intra16x16_pred_mode](dest_y, block_offset, h->mb + (p*256 << pixel_shift), linesize); - }else{ - for(i=0; i<16; i++){ - if(h->non_zero_count_cache[ scan8[i+p*16] ] || dctcoef_get(h->mb, pixel_shift, i*16+p*256)) - s->dsp.add_pixels4(dest_y + block_offset[i], h->mb + (i*16+p*256 << pixel_shift), linesize); - } + block_offset += 16 * p; + if (!IS_INTRA4x4(mb_type)) { + if (is_h264) { + if (IS_INTRA16x16(mb_type)) { + if (transform_bypass) { + if (h->sps.profile_idc == 244 && + (h->intra16x16_pred_mode == VERT_PRED8x8 || + h->intra16x16_pred_mode == HOR_PRED8x8)) { + h->hpc.pred16x16_add[h->intra16x16_pred_mode](dest_y, block_offset, + h->mb + (p * 256 << pixel_shift), + linesize); + } else { + for (i = 0; i < 16; i++) + if (h->non_zero_count_cache[scan8[i + p * 16]] || + dctcoef_get(h->mb, pixel_shift, i * 16 + p * 256)) + s->dsp.add_pixels4(dest_y + block_offset[i], + h->mb + (i * 16 + p * 256 << pixel_shift), + linesize); } - }else{ - h->h264dsp.h264_idct_add16intra(dest_y, block_offset, h->mb + (p*256 << pixel_shift), linesize, h->non_zero_count_cache+p*5*8); + } else { + h->h264dsp.h264_idct_add16intra(dest_y, block_offset, + h->mb + (p * 256 << pixel_shift), + linesize, + h->non_zero_count_cache + p * 5 * 8); } - }else if(h->cbp&15){ - if(transform_bypass){ + } else if (h->cbp & 15) { + if (transform_bypass) { const int di = IS_8x8DCT(mb_type) ? 4 : 1; - idct_add= IS_8x8DCT(mb_type) ? s->dsp.add_pixels8 : s->dsp.add_pixels4; - for(i=0; i<16; i+=di){ - if(h->non_zero_count_cache[ scan8[i+p*16] ]){ - idct_add(dest_y + block_offset[i], h->mb + (i*16+p*256 << pixel_shift), linesize); - } - } - }else{ - if(IS_8x8DCT(mb_type)){ - h->h264dsp.h264_idct8_add4(dest_y, block_offset, h->mb + (p*256 << pixel_shift), linesize, h->non_zero_count_cache+p*5*8); - }else{ - h->h264dsp.h264_idct_add16(dest_y, block_offset, h->mb + (p*256 << pixel_shift), linesize, h->non_zero_count_cache+p*5*8); - } + idct_add = IS_8x8DCT(mb_type) ? s->dsp.add_pixels8 + : s->dsp.add_pixels4; + for (i = 0; i < 16; i += di) + if (h->non_zero_count_cache[scan8[i + p * 16]]) + idct_add(dest_y + block_offset[i], + h->mb + (i * 16 + p * 256 << pixel_shift), + linesize); + } else { + if (IS_8x8DCT(mb_type)) + h->h264dsp.h264_idct8_add4(dest_y, block_offset, + h->mb + (p * 256 << pixel_shift), + linesize, + h->non_zero_count_cache + p * 5 * 8); + else + h->h264dsp.h264_idct_add16(dest_y, block_offset, + h->mb + (p * 256 << pixel_shift), + linesize, + h->non_zero_count_cache + p * 5 * 8); } } } else if (CONFIG_SVQ3_DECODER) { - for(i=0; i<16; i++){ - if(h->non_zero_count_cache[ scan8[i+p*16] ] || h->mb[i*16+p*256]){ //FIXME benchmark weird rule, & below - uint8_t * const ptr= dest_y + block_offset[i]; - ff_svq3_add_idct_c(ptr, h->mb + i*16 + p*256, linesize, s->qscale, IS_INTRA(mb_type) ? 1 : 0); + for (i = 0; i < 16; i++) + if (h->non_zero_count_cache[scan8[i + p * 16]] || h->mb[i * 16 + p * 256]) { + // FIXME benchmark weird rule, & below + uint8_t *const ptr = dest_y + block_offset[i]; + ff_svq3_add_idct_c(ptr, h->mb + i * 16 + p * 256, linesize, + s->qscale, IS_INTRA(mb_type) ? 1 : 0); } - } } } } -static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, int pixel_shift) +static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, + int pixel_shift) { - MpegEncContext * const s = &h->s; - const int mb_x= s->mb_x; - const int mb_y= s->mb_y; - const int mb_xy= h->mb_xy; + MpegEncContext *const s = &h->s; + const int mb_x = s->mb_x; + const int mb_y = s->mb_y; + const int mb_xy = h->mb_xy; const int mb_type = s->current_picture.f.mb_type[mb_xy]; - uint8_t *dest_y, *dest_cb, *dest_cr; + uint8_t *dest_y, *dest_cb, *dest_cr; int linesize, uvlinesize /*dct_offset*/; int i, j; int *block_offset = &h->block_offset[0]; @@ -1907,40 +2091,41 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i /* is_h264 should always be true if SVQ3 is disabled. */ const int is_h264 = !CONFIG_SVQ3_DECODER || simple || s->codec_id == CODEC_ID_H264; void (*idct_add)(uint8_t *dst, DCTELEM *block, int stride); - const int block_h = 16 >> s->chroma_y_shift; + const int block_h = 16 >> s->chroma_y_shift; const int chroma422 = CHROMA422; - dest_y = s->current_picture.f.data[0] + ((mb_x << pixel_shift) + mb_y * s->linesize ) * 16; - dest_cb = s->current_picture.f.data[1] + (mb_x << pixel_shift)*8 + mb_y * s->uvlinesize * block_h; - dest_cr = s->current_picture.f.data[2] + (mb_x << pixel_shift)*8 + mb_y * s->uvlinesize * block_h; + dest_y = s->current_picture.f.data[0] + ((mb_x << pixel_shift) + mb_y * s->linesize) * 16; + dest_cb = s->current_picture.f.data[1] + (mb_x << pixel_shift) * 8 + mb_y * s->uvlinesize * block_h; + dest_cr = s->current_picture.f.data[2] + (mb_x << pixel_shift) * 8 + mb_y * s->uvlinesize * block_h; - s->dsp.prefetch(dest_y + (s->mb_x&3)*4*s->linesize + (64 << pixel_shift), s->linesize, 4); - s->dsp.prefetch(dest_cb + (s->mb_x&7)*s->uvlinesize + (64 << pixel_shift), dest_cr - dest_cb, 2); + s->dsp.prefetch(dest_y + (s->mb_x & 3) * 4 * s->linesize + (64 << pixel_shift), s->linesize, 4); + s->dsp.prefetch(dest_cb + (s->mb_x & 7) * s->uvlinesize + (64 << pixel_shift), dest_cr - dest_cb, 2); - h->list_counts[mb_xy]= h->list_count; + h->list_counts[mb_xy] = h->list_count; if (!simple && MB_FIELD) { - linesize = h->mb_linesize = s->linesize * 2; - uvlinesize = h->mb_uvlinesize = s->uvlinesize * 2; + linesize = h->mb_linesize = s->linesize * 2; + uvlinesize = h->mb_uvlinesize = s->uvlinesize * 2; block_offset = &h->block_offset[48]; - if(mb_y&1){ //FIXME move out of this function? - dest_y -= s->linesize*15; - dest_cb-= s->uvlinesize * (block_h - 1); - dest_cr-= s->uvlinesize * (block_h - 1); + if (mb_y & 1) { // FIXME move out of this function? + dest_y -= s->linesize * 15; + dest_cb -= s->uvlinesize * (block_h - 1); + dest_cr -= s->uvlinesize * (block_h - 1); } - if(FRAME_MBAFF) { + if (FRAME_MBAFF) { int list; - for(list=0; listlist_count; list++){ - if(!USES_LIST(mb_type, list)) + for (list = 0; list < h->list_count; list++) { + if (!USES_LIST(mb_type, list)) continue; - if(IS_16X16(mb_type)){ + if (IS_16X16(mb_type)) { int8_t *ref = &h->ref_cache[list][scan8[0]]; - fill_rectangle(ref, 4, 4, 8, (16+*ref)^(s->mb_y&1), 1); - }else{ - for(i=0; i<16; i+=4){ + fill_rectangle(ref, 4, 4, 8, (16 + *ref) ^ (s->mb_y & 1), 1); + } else { + for (i = 0; i < 16; i += 4) { int ref = h->ref_cache[list][scan8[i]]; - if(ref >= 0) - fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2, 8, (16+ref)^(s->mb_y&1), 1); + if (ref >= 0) + fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2, + 8, (16 + ref) ^ (s->mb_y & 1), 1); } } } @@ -1948,7 +2133,7 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i } else { linesize = h->mb_linesize = s->linesize; uvlinesize = h->mb_uvlinesize = s->uvlinesize; -// dct_offset = s->linesize * 16; + // dct_offset = s->linesize * 16; } if (!simple && IS_INTRA_PCM(mb_type)) { @@ -1956,73 +2141,74 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i const int bit_depth = h->sps.bit_depth_luma; int j; GetBitContext gb; - init_get_bits(&gb, (uint8_t*)h->mb, 384*bit_depth); + init_get_bits(&gb, (uint8_t *)h->mb, 384 * bit_depth); for (i = 0; i < 16; i++) { - uint16_t *tmp_y = (uint16_t*)(dest_y + i*linesize); + uint16_t *tmp_y = (uint16_t *)(dest_y + i * linesize); for (j = 0; j < 16; j++) tmp_y[j] = get_bits(&gb, bit_depth); } - if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){ + if (simple || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) { 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++) { + 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++) { + uint16_t *tmp_cr = (uint16_t *)(dest_cr + i * uvlinesize); + for (j = 0; j < 8; j++) tmp_cr[j] = 1 << (bit_depth - 1); - } } } else { for (i = 0; i < block_h; i++) { - uint16_t *tmp_cb = (uint16_t*)(dest_cb + i*uvlinesize); + uint16_t *tmp_cb = (uint16_t *)(dest_cb + i * uvlinesize); for (j = 0; j < 8; j++) tmp_cb[j] = get_bits(&gb, bit_depth); } for (i = 0; i < block_h; i++) { - uint16_t *tmp_cr = (uint16_t*)(dest_cr + i*uvlinesize); + uint16_t *tmp_cr = (uint16_t *)(dest_cr + i * uvlinesize); for (j = 0; j < 8; j++) tmp_cr[j] = get_bits(&gb, bit_depth); } } } } else { - for (i=0; i<16; i++) { - memcpy(dest_y + i* linesize, h->mb + i*8, 16); - } - if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){ + for (i = 0; i < 16; i++) + memcpy(dest_y + i * linesize, h->mb + i * 8, 16); + 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); + memset(dest_cb + i * uvlinesize, 128, 8); + memset(dest_cr + i * uvlinesize, 128, 8); } } else { 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); + memcpy(dest_cb + i * uvlinesize, h->mb + 128 + i * 4, 8); + memcpy(dest_cr + i * uvlinesize, h->mb + 160 + i * 4, 8); } } } } } else { - if(IS_INTRA(mb_type)){ - if(h->deblocking_filter) - xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 1, 0, simple, pixel_shift); + if (IS_INTRA(mb_type)) { + if (h->deblocking_filter) + xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, + uvlinesize, 1, 0, simple, pixel_shift); - if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){ - h->hpc.pred8x8[ h->chroma_pred_mode ](dest_cb, uvlinesize); - h->hpc.pred8x8[ h->chroma_pred_mode ](dest_cr, uvlinesize); + if (simple || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) { + h->hpc.pred8x8[h->chroma_pred_mode](dest_cb, uvlinesize); + h->hpc.pred8x8[h->chroma_pred_mode](dest_cr, uvlinesize); } - hl_decode_mb_predict_luma(h, mb_type, is_h264, simple, transform_bypass, pixel_shift, block_offset, linesize, dest_y, 0); + hl_decode_mb_predict_luma(h, mb_type, is_h264, simple, + transform_bypass, pixel_shift, + block_offset, linesize, dest_y, 0); - if(h->deblocking_filter) - xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0, 0, simple, pixel_shift); - }else if(is_h264){ + if (h->deblocking_filter) + xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, + uvlinesize, 0, 0, simple, pixel_shift); + } else if (is_h264) { if (chroma422) { hl_motion_422(h, dest_y, dest_cb, dest_cr, s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab, @@ -2040,31 +2226,45 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i } } - hl_decode_mb_idct_luma(h, mb_type, is_h264, simple, transform_bypass, pixel_shift, block_offset, linesize, dest_y, 0); + hl_decode_mb_idct_luma(h, mb_type, is_h264, simple, transform_bypass, + pixel_shift, block_offset, linesize, dest_y, 0); - if((simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)) && (h->cbp&0x30)){ - uint8_t *dest[2] = {dest_cb, dest_cr}; - if(transform_bypass){ - if(IS_INTRA(mb_type) && h->sps.profile_idc==244 && (h->chroma_pred_mode==VERT_PRED8x8 || h->chroma_pred_mode==HOR_PRED8x8)){ - h->hpc.pred8x8_add[h->chroma_pred_mode](dest[0], block_offset + 16, h->mb + (16*16*1 << pixel_shift), uvlinesize); - h->hpc.pred8x8_add[h->chroma_pred_mode](dest[1], block_offset + 32, h->mb + (16*16*2 << pixel_shift), uvlinesize); - }else{ + if ((simple || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) && + (h->cbp & 0x30)) { + uint8_t *dest[2] = { dest_cb, dest_cr }; + if (transform_bypass) { + if (IS_INTRA(mb_type) && h->sps.profile_idc == 244 && + (h->chroma_pred_mode == VERT_PRED8x8 || + h->chroma_pred_mode == HOR_PRED8x8)) { + h->hpc.pred8x8_add[h->chroma_pred_mode](dest[0], + block_offset + 16, + h->mb + (16 * 16 * 1 << pixel_shift), + uvlinesize); + h->hpc.pred8x8_add[h->chroma_pred_mode](dest[1], + block_offset + 32, + h->mb + (16 * 16 * 2 << pixel_shift), + uvlinesize); + } else { idct_add = s->dsp.add_pixels4; - for(j=1; j<3; j++){ - for(i=j*16; inon_zero_count_cache[ scan8[i] ] || dctcoef_get(h->mb, pixel_shift, i*16)) - idct_add (dest[j-1] + block_offset[i], h->mb + (i*16 << pixel_shift), uvlinesize); - } + for (j = 1; j < 3; j++) { + for (i = j * 16; i < j * 16 + 4; i++) + if (h->non_zero_count_cache[scan8[i]] || + dctcoef_get(h->mb, pixel_shift, i * 16)) + idct_add(dest[j - 1] + block_offset[i], + h->mb + (i * 16 << pixel_shift), + uvlinesize); if (chroma422) { - for(i=j*16+4; inon_zero_count_cache[ scan8[i+4] ] || dctcoef_get(h->mb, pixel_shift, i*16)) - idct_add (dest[j-1] + block_offset[i+4], h->mb + (i*16 << pixel_shift), uvlinesize); - } + for (i = j * 16 + 4; i < j * 16 + 8; i++) + if (h->non_zero_count_cache[scan8[i + 4]] || + dctcoef_get(h->mb, pixel_shift, i * 16)) + idct_add(dest[j - 1] + block_offset[i + 4], + h->mb + (i * 16 << pixel_shift), + uvlinesize); } } } - }else{ - if(is_h264){ + } else { + if (is_h264) { int qp[2]; if (chroma422) { qp[0] = h->chroma_qp[0] + 3; @@ -2073,114 +2273,124 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i qp[0] = h->chroma_qp[0]; qp[1] = h->chroma_qp[1]; } - if(h->non_zero_count_cache[ scan8[CHROMA_DC_BLOCK_INDEX+0] ]) - h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16*16*1 << pixel_shift), h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][qp[0]][0]); - if(h->non_zero_count_cache[ scan8[CHROMA_DC_BLOCK_INDEX+1] ]) - h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16*16*2 << pixel_shift), h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][qp[1]][0]); + if (h->non_zero_count_cache[scan8[CHROMA_DC_BLOCK_INDEX + 0]]) + h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16 * 16 * 1 << pixel_shift), + h->dequant4_coeff[IS_INTRA(mb_type) ? 1 : 4][qp[0]][0]); + if (h->non_zero_count_cache[scan8[CHROMA_DC_BLOCK_INDEX + 1]]) + h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16 * 16 * 2 << pixel_shift), + h->dequant4_coeff[IS_INTRA(mb_type) ? 2 : 5][qp[1]][0]); h->h264dsp.h264_idct_add8(dest, block_offset, h->mb, uvlinesize, h->non_zero_count_cache); } else if (CONFIG_SVQ3_DECODER) { - 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++){ - for(i=j*16; inon_zero_count_cache[ scan8[i] ] || h->mb[i*16]){ - uint8_t * const ptr= dest[j-1] + block_offset[i]; - ff_svq3_add_idct_c(ptr, h->mb + i*16, uvlinesize, ff_h264_chroma_qp[0][s->qscale + 12] - 12, 2); + 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++) { + for (i = j * 16; i < j * 16 + 4; i++) + if (h->non_zero_count_cache[scan8[i]] || h->mb[i * 16]) { + uint8_t *const ptr = dest[j - 1] + block_offset[i]; + ff_svq3_add_idct_c(ptr, h->mb + i * 16, + uvlinesize, + ff_h264_chroma_qp[0][s->qscale + 12] - 12, 2); } - } } } } } } - if(h->cbp || IS_INTRA(mb_type)) - { + if (h->cbp || IS_INTRA(mb_type)) { s->dsp.clear_blocks(h->mb); - s->dsp.clear_blocks(h->mb+(24*16<dsp.clear_blocks(h->mb + (24 * 16 << pixel_shift)); } } -static av_always_inline void hl_decode_mb_444_internal(H264Context *h, int simple, int pixel_shift){ - MpegEncContext * const s = &h->s; - const int mb_x= s->mb_x; - const int mb_y= s->mb_y; - const int mb_xy= h->mb_xy; +static av_always_inline void hl_decode_mb_444_internal(H264Context *h, + int simple, + int pixel_shift) +{ + MpegEncContext *const s = &h->s; + const int mb_x = s->mb_x; + const int mb_y = s->mb_y; + const int mb_xy = h->mb_xy; const int mb_type = s->current_picture.f.mb_type[mb_xy]; - uint8_t *dest[3]; + uint8_t *dest[3]; int linesize; int i, j, p; int *block_offset = &h->block_offset[0]; const int transform_bypass = !simple && (s->qscale == 0 && h->sps.transform_bypass); - const int plane_count = (simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)) ? 3 : 1; + const int plane_count = (simple || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) ? 3 : 1; - for (p = 0; p < plane_count; p++) - { - dest[p] = s->current_picture.f.data[p] + ((mb_x << pixel_shift) + mb_y * s->linesize) * 16; - s->dsp.prefetch(dest[p] + (s->mb_x&3)*4*s->linesize + (64 << pixel_shift), s->linesize, 4); + for (p = 0; p < plane_count; p++) { + dest[p] = s->current_picture.f.data[p] + + ((mb_x << pixel_shift) + mb_y * s->linesize) * 16; + s->dsp.prefetch(dest[p] + (s->mb_x & 3) * 4 * s->linesize + (64 << pixel_shift), + s->linesize, 4); } - h->list_counts[mb_xy]= h->list_count; + h->list_counts[mb_xy] = h->list_count; if (!simple && MB_FIELD) { - linesize = h->mb_linesize = h->mb_uvlinesize = s->linesize * 2; + linesize = h->mb_linesize = h->mb_uvlinesize = s->linesize * 2; block_offset = &h->block_offset[48]; - if(mb_y&1) //FIXME move out of this function? + if (mb_y & 1) // FIXME move out of this function? for (p = 0; p < 3; p++) - dest[p] -= s->linesize*15; - if(FRAME_MBAFF) { + dest[p] -= s->linesize * 15; + if (FRAME_MBAFF) { int list; - for(list=0; listlist_count; list++){ - if(!USES_LIST(mb_type, list)) + for (list = 0; list < h->list_count; list++) { + if (!USES_LIST(mb_type, list)) continue; - if(IS_16X16(mb_type)){ + if (IS_16X16(mb_type)) { int8_t *ref = &h->ref_cache[list][scan8[0]]; - fill_rectangle(ref, 4, 4, 8, (16+*ref)^(s->mb_y&1), 1); - }else{ - for(i=0; i<16; i+=4){ + fill_rectangle(ref, 4, 4, 8, (16 + *ref) ^ (s->mb_y & 1), 1); + } else { + for (i = 0; i < 16; i += 4) { int ref = h->ref_cache[list][scan8[i]]; - if(ref >= 0) - fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2, 8, (16+ref)^(s->mb_y&1), 1); + if (ref >= 0) + fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2, + 8, (16 + ref) ^ (s->mb_y & 1), 1); } } } } } else { - linesize = h->mb_linesize = h->mb_uvlinesize = s->linesize; + linesize = h->mb_linesize = h->mb_uvlinesize = s->linesize; } if (!simple && IS_INTRA_PCM(mb_type)) { if (pixel_shift) { const int bit_depth = h->sps.bit_depth_luma; GetBitContext gb; - init_get_bits(&gb, (uint8_t*)h->mb, 768*bit_depth); + init_get_bits(&gb, (uint8_t *)h->mb, 768 * bit_depth); - for (p = 0; p < plane_count; p++) { + for (p = 0; p < plane_count; p++) for (i = 0; i < 16; i++) { - uint16_t *tmp = (uint16_t*)(dest[p] + i*linesize); + uint16_t *tmp = (uint16_t *)(dest[p] + i * linesize); for (j = 0; j < 16; j++) tmp[j] = get_bits(&gb, bit_depth); } - } } else { - for (p = 0; p < plane_count; p++) { - for (i = 0; i < 16; i++) { - memcpy(dest[p] + i*linesize, h->mb + p*128 + i*8, 16); - } - } + for (p = 0; p < plane_count; p++) + for (i = 0; i < 16; i++) + memcpy(dest[p] + i * linesize, h->mb + p * 128 + i * 8, 16); } } else { - if(IS_INTRA(mb_type)){ - if(h->deblocking_filter) - xchg_mb_border(h, dest[0], dest[1], dest[2], linesize, linesize, 1, 1, simple, pixel_shift); + if (IS_INTRA(mb_type)) { + if (h->deblocking_filter) + xchg_mb_border(h, dest[0], dest[1], dest[2], linesize, + linesize, 1, 1, simple, pixel_shift); for (p = 0; p < plane_count; p++) - hl_decode_mb_predict_luma(h, mb_type, 1, simple, transform_bypass, pixel_shift, block_offset, linesize, dest[p], p); + hl_decode_mb_predict_luma(h, mb_type, 1, simple, + transform_bypass, pixel_shift, + block_offset, linesize, dest[p], p); - if(h->deblocking_filter) - xchg_mb_border(h, dest[0], dest[1], dest[2], linesize, linesize, 0, 1, simple, pixel_shift); - }else{ + if (h->deblocking_filter) + xchg_mb_border(h, dest[0], dest[1], dest[2], linesize, + linesize, 0, 1, simple, pixel_shift); + } else { hl_motion(h, dest[0], dest[1], dest[2], s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab, s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab, @@ -2189,48 +2399,55 @@ static av_always_inline void hl_decode_mb_444_internal(H264Context *h, int simpl } for (p = 0; p < plane_count; p++) - hl_decode_mb_idct_luma(h, mb_type, 1, simple, transform_bypass, pixel_shift, block_offset, linesize, dest[p], p); + hl_decode_mb_idct_luma(h, mb_type, 1, simple, transform_bypass, + pixel_shift, block_offset, linesize, + dest[p], p); } - if(h->cbp || IS_INTRA(mb_type)) - { + if (h->cbp || IS_INTRA(mb_type)) { s->dsp.clear_blocks(h->mb); - s->dsp.clear_blocks(h->mb+(24*16<dsp.clear_blocks(h->mb + (24 * 16 << pixel_shift)); } } /** * Process a macroblock; this case avoids checks for expensive uncommon cases. */ -#define hl_decode_mb_simple(sh, bits) \ -static void hl_decode_mb_simple_ ## bits(H264Context *h){ \ - hl_decode_mb_internal(h, 1, sh); \ +#define hl_decode_mb_simple(sh, bits) \ +static void hl_decode_mb_simple_ ## bits(H264Context *h) \ +{ \ + hl_decode_mb_internal(h, 1, sh); \ } + hl_decode_mb_simple(0, 8) hl_decode_mb_simple(1, 16) /** * Process a macroblock; this handles edge cases, such as interlacing. */ -static void av_noinline hl_decode_mb_complex(H264Context *h){ +static void av_noinline hl_decode_mb_complex(H264Context *h) +{ hl_decode_mb_internal(h, 0, h->pixel_shift); } -static void av_noinline hl_decode_mb_444_complex(H264Context *h){ +static void av_noinline hl_decode_mb_444_complex(H264Context *h) +{ hl_decode_mb_444_internal(h, 0, h->pixel_shift); } -static void av_noinline hl_decode_mb_444_simple(H264Context *h){ +static void av_noinline hl_decode_mb_444_simple(H264Context *h) +{ hl_decode_mb_444_internal(h, 1, 0); } -void ff_h264_hl_decode_mb(H264Context *h){ - MpegEncContext * const s = &h->s; - const int mb_xy= h->mb_xy; +void ff_h264_hl_decode_mb(H264Context *h) +{ + MpegEncContext *const s = &h->s; + const int mb_xy = h->mb_xy; const int mb_type = s->current_picture.f.mb_type[mb_xy]; - int is_complex = CONFIG_SMALL || h->is_complex || IS_INTRA_PCM(mb_type) || s->qscale == 0; + int is_complex = CONFIG_SMALL || h->is_complex || IS_INTRA_PCM(mb_type) || s->qscale == 0; if (CHROMA444) { - if(is_complex || h->pixel_shift) + if (is_complex || h->pixel_shift) hl_decode_mb_444_complex(h); else hl_decode_mb_444_simple(h); @@ -2242,64 +2459,66 @@ void ff_h264_hl_decode_mb(H264Context *h){ hl_decode_mb_simple_8(h); } -static int pred_weight_table(H264Context *h){ - MpegEncContext * const s = &h->s; +static int pred_weight_table(H264Context *h) +{ + MpegEncContext *const s = &h->s; int list, i; int luma_def, chroma_def; - h->use_weight= 0; - h->use_weight_chroma= 0; - h->luma_log2_weight_denom= get_ue_golomb(&s->gb); - if(h->sps.chroma_format_idc) - h->chroma_log2_weight_denom= get_ue_golomb(&s->gb); - luma_def = 1<luma_log2_weight_denom; - chroma_def = 1<chroma_log2_weight_denom; + h->use_weight = 0; + h->use_weight_chroma = 0; + h->luma_log2_weight_denom = get_ue_golomb(&s->gb); + if (h->sps.chroma_format_idc) + h->chroma_log2_weight_denom = get_ue_golomb(&s->gb); + luma_def = 1 << h->luma_log2_weight_denom; + chroma_def = 1 << h->chroma_log2_weight_denom; - for(list=0; list<2; list++){ + for (list = 0; list < 2; list++) { h->luma_weight_flag[list] = 0; h->chroma_weight_flag[list] = 0; - for(i=0; iref_count[list]; i++){ + for (i = 0; i < h->ref_count[list]; i++) { int luma_weight_flag, chroma_weight_flag; - luma_weight_flag= get_bits1(&s->gb); - if(luma_weight_flag){ - h->luma_weight[i][list][0]= get_se_golomb(&s->gb); - h->luma_weight[i][list][1]= get_se_golomb(&s->gb); - if( h->luma_weight[i][list][0] != luma_def - || h->luma_weight[i][list][1] != 0) { - h->use_weight= 1; - h->luma_weight_flag[list]= 1; + luma_weight_flag = get_bits1(&s->gb); + if (luma_weight_flag) { + h->luma_weight[i][list][0] = get_se_golomb(&s->gb); + h->luma_weight[i][list][1] = get_se_golomb(&s->gb); + if (h->luma_weight[i][list][0] != luma_def || + h->luma_weight[i][list][1] != 0) { + h->use_weight = 1; + h->luma_weight_flag[list] = 1; } - }else{ - h->luma_weight[i][list][0]= luma_def; - h->luma_weight[i][list][1]= 0; + } else { + h->luma_weight[i][list][0] = luma_def; + h->luma_weight[i][list][1] = 0; } - if(h->sps.chroma_format_idc){ - chroma_weight_flag= get_bits1(&s->gb); - if(chroma_weight_flag){ + if (h->sps.chroma_format_idc) { + chroma_weight_flag = get_bits1(&s->gb); + if (chroma_weight_flag) { int j; - for(j=0; j<2; j++){ - h->chroma_weight[i][list][j][0]= get_se_golomb(&s->gb); - h->chroma_weight[i][list][j][1]= get_se_golomb(&s->gb); - if( h->chroma_weight[i][list][j][0] != chroma_def - || h->chroma_weight[i][list][j][1] != 0) { - h->use_weight_chroma= 1; - h->chroma_weight_flag[list]= 1; + for (j = 0; j < 2; j++) { + h->chroma_weight[i][list][j][0] = get_se_golomb(&s->gb); + h->chroma_weight[i][list][j][1] = get_se_golomb(&s->gb); + if (h->chroma_weight[i][list][j][0] != chroma_def || + h->chroma_weight[i][list][j][1] != 0) { + h->use_weight_chroma = 1; + h->chroma_weight_flag[list] = 1; } } - }else{ + } else { int j; - for(j=0; j<2; j++){ - h->chroma_weight[i][list][j][0]= chroma_def; - h->chroma_weight[i][list][j][1]= 0; + for (j = 0; j < 2; j++) { + h->chroma_weight[i][list][j][0] = chroma_def; + h->chroma_weight[i][list][j][1] = 0; } } } } - if(h->slice_type_nos != AV_PICTURE_TYPE_B) break; + if (h->slice_type_nos != AV_PICTURE_TYPE_B) + break; } - h->use_weight= h->use_weight || h->use_weight_chroma; + h->use_weight = h->use_weight || h->use_weight_chroma; return 0; } @@ -2308,8 +2527,9 @@ static int pred_weight_table(H264Context *h){ * @param field 0/1 initialize the weight for interlaced MBAFF * -1 initializes the rest */ -static void implicit_weight_table(H264Context *h, int field){ - MpegEncContext * const s = &h->s; +static void implicit_weight_table(H264Context *h, int field) +{ + MpegEncContext *const s = &h->s; int ref0, ref1, i, cur_poc, ref_start, ref_count0, ref_count1; for (i = 0; i < 2; i++) { @@ -2317,53 +2537,53 @@ static void implicit_weight_table(H264Context *h, int field){ h->chroma_weight_flag[i] = 0; } - if(field < 0){ + if (field < 0) { if (s->picture_structure == PICT_FRAME) { cur_poc = s->current_picture_ptr->poc; } else { cur_poc = s->current_picture_ptr->field_poc[s->picture_structure - 1]; } - if( h->ref_count[0] == 1 && h->ref_count[1] == 1 && !FRAME_MBAFF - && h->ref_list[0][0].poc + h->ref_list[1][0].poc == 2*cur_poc){ - h->use_weight= 0; - h->use_weight_chroma= 0; - return; - } - ref_start= 0; - ref_count0= h->ref_count[0]; - ref_count1= h->ref_count[1]; - }else{ - cur_poc = s->current_picture_ptr->field_poc[field]; - ref_start= 16; - ref_count0= 16+2*h->ref_count[0]; - ref_count1= 16+2*h->ref_count[1]; + if (h->ref_count[0] == 1 && h->ref_count[1] == 1 && !FRAME_MBAFF && + h->ref_list[0][0].poc + h->ref_list[1][0].poc == 2 * cur_poc) { + h->use_weight = 0; + h->use_weight_chroma = 0; + return; + } + ref_start = 0; + ref_count0 = h->ref_count[0]; + ref_count1 = h->ref_count[1]; + } else { + cur_poc = s->current_picture_ptr->field_poc[field]; + ref_start = 16; + ref_count0 = 16 + 2 * h->ref_count[0]; + ref_count1 = 16 + 2 * h->ref_count[1]; } - h->use_weight= 2; - h->use_weight_chroma= 2; - h->luma_log2_weight_denom= 5; - h->chroma_log2_weight_denom= 5; + h->use_weight = 2; + h->use_weight_chroma = 2; + h->luma_log2_weight_denom = 5; + h->chroma_log2_weight_denom = 5; - for(ref0=ref_start; ref0 < ref_count0; ref0++){ + for (ref0 = ref_start; ref0 < ref_count0; ref0++) { int poc0 = h->ref_list[0][ref0].poc; - for(ref1=ref_start; ref1 < ref_count1; ref1++){ + for (ref1 = ref_start; ref1 < ref_count1; ref1++) { int w = 32; if (!h->ref_list[0][ref0].long_ref && !h->ref_list[1][ref1].long_ref) { int poc1 = h->ref_list[1][ref1].poc; - int td = av_clip(poc1 - poc0, -128, 127); - if(td){ + int td = av_clip(poc1 - poc0, -128, 127); + if (td) { int tb = av_clip(cur_poc - poc0, -128, 127); int tx = (16384 + (FFABS(td) >> 1)) / td; - int dist_scale_factor = (tb*tx + 32) >> 8; - if(dist_scale_factor >= -64 && dist_scale_factor <= 128) + int dist_scale_factor = (tb * tx + 32) >> 8; + if (dist_scale_factor >= -64 && dist_scale_factor <= 128) w = 64 - dist_scale_factor; } } - if(field<0){ - h->implicit_weight[ref0][ref1][0]= - h->implicit_weight[ref0][ref1][1]= w; - }else{ - h->implicit_weight[ref0][ref1][field]=w; + if (field < 0) { + h->implicit_weight[ref0][ref1][0] = + h->implicit_weight[ref0][ref1][1] = w; + } else { + h->implicit_weight[ref0][ref1][field] = w; } } } @@ -2372,140 +2592,144 @@ static void implicit_weight_table(H264Context *h, int field){ /** * instantaneous decoder refresh. */ -static void idr(H264Context *h){ +static void idr(H264Context *h) +{ ff_h264_remove_all_refs(h); - h->prev_frame_num= 0; - h->prev_frame_num_offset= 0; - h->prev_poc_msb= - h->prev_poc_lsb= 0; + h->prev_frame_num = 0; + h->prev_frame_num_offset = 0; + h->prev_poc_msb = + h->prev_poc_lsb = 0; } /* forget old pics after a seek */ -static void flush_dpb(AVCodecContext *avctx){ - H264Context *h= avctx->priv_data; +static void flush_dpb(AVCodecContext *avctx) +{ + H264Context *h = avctx->priv_data; int i; - for(i=0; idelayed_pic[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->delayed_pic[i] = NULL; } for (i = 0; i < MAX_DELAYED_PIC_COUNT; i++) h->last_pocs[i] = INT_MIN; - h->outputed_poc=h->next_outputed_poc= INT_MIN; + h->outputed_poc = h->next_outputed_poc = INT_MIN; h->prev_interlaced_frame = 1; idr(h); - if(h->s.current_picture_ptr) + if (h->s.current_picture_ptr) h->s.current_picture_ptr->f.reference = 0; - h->s.first_field= 0; + h->s.first_field = 0; ff_h264_reset_sei(h); ff_mpeg_flush(avctx); } -static int init_poc(H264Context *h){ - MpegEncContext * const s = &h->s; - const int max_frame_num= 1<sps.log2_max_frame_num; +static int init_poc(H264Context *h) +{ + MpegEncContext *const s = &h->s; + const int max_frame_num = 1 << h->sps.log2_max_frame_num; int field_poc[2]; Picture *cur = s->current_picture_ptr; - h->frame_num_offset= h->prev_frame_num_offset; - if(h->frame_num < h->prev_frame_num) + h->frame_num_offset = h->prev_frame_num_offset; + if (h->frame_num < h->prev_frame_num) h->frame_num_offset += max_frame_num; - if(h->sps.poc_type==0){ - const int max_poc_lsb= 1<sps.log2_max_poc_lsb; + if (h->sps.poc_type == 0) { + const int max_poc_lsb = 1 << h->sps.log2_max_poc_lsb; - if (h->poc_lsb < h->prev_poc_lsb && h->prev_poc_lsb - h->poc_lsb >= max_poc_lsb/2) + if (h->poc_lsb < h->prev_poc_lsb && h->prev_poc_lsb - h->poc_lsb >= max_poc_lsb / 2) h->poc_msb = h->prev_poc_msb + max_poc_lsb; - else if(h->poc_lsb > h->prev_poc_lsb && h->prev_poc_lsb - h->poc_lsb < -max_poc_lsb/2) + else if (h->poc_lsb > h->prev_poc_lsb && h->prev_poc_lsb - h->poc_lsb < -max_poc_lsb / 2) h->poc_msb = h->prev_poc_msb - max_poc_lsb; else h->poc_msb = h->prev_poc_msb; -//printf("poc: %d %d\n", h->poc_msb, h->poc_lsb); + // printf("poc: %d %d\n", h->poc_msb, h->poc_lsb); field_poc[0] = field_poc[1] = h->poc_msb + h->poc_lsb; - if(s->picture_structure == PICT_FRAME) + if (s->picture_structure == PICT_FRAME) field_poc[1] += h->delta_poc_bottom; - }else if(h->sps.poc_type==1){ + } else if (h->sps.poc_type == 1) { int abs_frame_num, expected_delta_per_poc_cycle, expectedpoc; int i; - if(h->sps.poc_cycle_length != 0) + if (h->sps.poc_cycle_length != 0) abs_frame_num = h->frame_num_offset + h->frame_num; else abs_frame_num = 0; - if(h->nal_ref_idc==0 && abs_frame_num > 0) + if (h->nal_ref_idc == 0 && abs_frame_num > 0) abs_frame_num--; expected_delta_per_poc_cycle = 0; - for(i=0; i < h->sps.poc_cycle_length; i++) - expected_delta_per_poc_cycle += h->sps.offset_for_ref_frame[ i ]; //FIXME integrate during sps parse + for (i = 0; i < h->sps.poc_cycle_length; i++) + // FIXME integrate during sps parse + expected_delta_per_poc_cycle += h->sps.offset_for_ref_frame[i]; - if(abs_frame_num > 0){ + if (abs_frame_num > 0) { int poc_cycle_cnt = (abs_frame_num - 1) / h->sps.poc_cycle_length; int frame_num_in_poc_cycle = (abs_frame_num - 1) % h->sps.poc_cycle_length; expectedpoc = poc_cycle_cnt * expected_delta_per_poc_cycle; - for(i = 0; i <= frame_num_in_poc_cycle; i++) - expectedpoc = expectedpoc + h->sps.offset_for_ref_frame[ i ]; + for (i = 0; i <= frame_num_in_poc_cycle; i++) + expectedpoc = expectedpoc + h->sps.offset_for_ref_frame[i]; } else expectedpoc = 0; - if(h->nal_ref_idc == 0) + if (h->nal_ref_idc == 0) expectedpoc = expectedpoc + h->sps.offset_for_non_ref_pic; field_poc[0] = expectedpoc + h->delta_poc[0]; field_poc[1] = field_poc[0] + h->sps.offset_for_top_to_bottom_field; - if(s->picture_structure == PICT_FRAME) + if (s->picture_structure == PICT_FRAME) field_poc[1] += h->delta_poc[1]; - }else{ - int poc= 2*(h->frame_num_offset + h->frame_num); + } else { + int poc = 2 * (h->frame_num_offset + h->frame_num); - if(!h->nal_ref_idc) + if (!h->nal_ref_idc) poc--; - field_poc[0]= poc; - field_poc[1]= poc; + field_poc[0] = poc; + field_poc[1] = poc; } - if(s->picture_structure != PICT_BOTTOM_FIELD) - s->current_picture_ptr->field_poc[0]= field_poc[0]; - if(s->picture_structure != PICT_TOP_FIELD) - s->current_picture_ptr->field_poc[1]= field_poc[1]; - cur->poc= FFMIN(cur->field_poc[0], cur->field_poc[1]); + if (s->picture_structure != PICT_BOTTOM_FIELD) + s->current_picture_ptr->field_poc[0] = field_poc[0]; + if (s->picture_structure != PICT_TOP_FIELD) + s->current_picture_ptr->field_poc[1] = field_poc[1]; + cur->poc = FFMIN(cur->field_poc[0], cur->field_poc[1]); return 0; } - /** * initialize scan tables */ -static void init_scan_tables(H264Context *h){ +static void init_scan_tables(H264Context *h) +{ int i; - for(i=0; i<16; i++){ -#define T(x) (x>>2) | ((x<<2) & 0xF) + for (i = 0; i < 16; i++) { +#define T(x) (x >> 2) | ((x << 2) & 0xF) h->zigzag_scan[i] = T(zigzag_scan[i]); - h-> field_scan[i] = T( field_scan[i]); + h->field_scan[i] = T(field_scan[i]); #undef T } - for(i=0; i<64; i++){ -#define T(x) (x>>3) | ((x&7)<<3) + for (i = 0; i < 64; i++) { +#define T(x) (x >> 3) | ((x & 7) << 3) h->zigzag_scan8x8[i] = T(ff_zigzag_direct[i]); h->zigzag_scan8x8_cavlc[i] = T(zigzag_scan8x8_cavlc[i]); h->field_scan8x8[i] = T(field_scan8x8[i]); h->field_scan8x8_cavlc[i] = T(field_scan8x8_cavlc[i]); #undef T } - if(h->sps.transform_bypass){ //FIXME same ugly + if (h->sps.transform_bypass) { // FIXME same ugly h->zigzag_scan_q0 = zigzag_scan; h->zigzag_scan8x8_q0 = ff_zigzag_direct; h->zigzag_scan8x8_cavlc_q0 = zigzag_scan8x8_cavlc; h->field_scan_q0 = field_scan; h->field_scan8x8_q0 = field_scan8x8; h->field_scan8x8_cavlc_q0 = field_scan8x8_cavlc; - }else{ + } else { h->zigzag_scan_q0 = h->zigzag_scan; h->zigzag_scan8x8_q0 = h->zigzag_scan8x8; h->zigzag_scan8x8_cavlc_q0 = h->zigzag_scan8x8_cavlc; @@ -2515,37 +2739,41 @@ static void init_scan_tables(H264Context *h){ } } -static int field_end(H264Context *h, int in_setup){ - MpegEncContext * const s = &h->s; - AVCodecContext * const avctx= s->avctx; +static int field_end(H264Context *h, int in_setup) +{ + MpegEncContext *const s = &h->s; + AVCodecContext *const avctx = s->avctx; int err = 0; - s->mb_y= 0; + s->mb_y = 0; if (!in_setup && !s->dropable) ff_thread_report_progress(&s->current_picture_ptr->f, (16 * s->mb_height >> FIELD_PICTURE) - 1, s->picture_structure == PICT_BOTTOM_FIELD); - if (CONFIG_H264_VDPAU_DECODER && s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU) + if (CONFIG_H264_VDPAU_DECODER && + s->avctx->codec->capabilities & CODEC_CAP_HWACCEL_VDPAU) ff_vdpau_h264_set_reference_frames(s); - if(in_setup || !(avctx->active_thread_type&FF_THREAD_FRAME)){ - if(!s->dropable) { + if (in_setup || !(avctx->active_thread_type & FF_THREAD_FRAME)) { + if (!s->dropable) { err = ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index); - h->prev_poc_msb= h->poc_msb; - h->prev_poc_lsb= h->poc_lsb; + h->prev_poc_msb = h->poc_msb; + h->prev_poc_lsb = h->poc_lsb; } - h->prev_frame_num_offset= h->frame_num_offset; - h->prev_frame_num= h->frame_num; - h->outputed_poc = h->next_outputed_poc; + h->prev_frame_num_offset = h->frame_num_offset; + h->prev_frame_num = h->frame_num; + h->outputed_poc = h->next_outputed_poc; } if (avctx->hwaccel) { if (avctx->hwaccel->end_frame(avctx) < 0) - av_log(avctx, AV_LOG_ERROR, "hardware accelerator failed to decode picture\n"); + av_log(avctx, AV_LOG_ERROR, + "hardware accelerator failed to decode picture\n"); } - if (CONFIG_H264_VDPAU_DECODER && s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU) + if (CONFIG_H264_VDPAU_DECODER && + s->avctx->codec->capabilities & CODEC_CAP_HWACCEL_VDPAU) ff_vdpau_h264_picture_complete(s); /* @@ -2565,7 +2793,7 @@ static int field_end(H264Context *h, int in_setup){ ff_MPV_frame_end(s); - h->current_slice=0; + h->current_slice = 0; return err; } @@ -2575,18 +2803,18 @@ static int field_end(H264Context *h, int in_setup){ */ static void clone_slice(H264Context *dst, H264Context *src) { - memcpy(dst->block_offset, src->block_offset, sizeof(dst->block_offset)); - dst->s.current_picture_ptr = src->s.current_picture_ptr; - dst->s.current_picture = src->s.current_picture; - dst->s.linesize = src->s.linesize; - dst->s.uvlinesize = src->s.uvlinesize; - dst->s.first_field = src->s.first_field; + memcpy(dst->block_offset, src->block_offset, sizeof(dst->block_offset)); + dst->s.current_picture_ptr = src->s.current_picture_ptr; + dst->s.current_picture = src->s.current_picture; + dst->s.linesize = src->s.linesize; + dst->s.uvlinesize = src->s.uvlinesize; + dst->s.first_field = src->s.first_field; - dst->prev_poc_msb = src->prev_poc_msb; - dst->prev_poc_lsb = src->prev_poc_lsb; - dst->prev_frame_num_offset = src->prev_frame_num_offset; - dst->prev_frame_num = src->prev_frame_num; - dst->short_ref_count = src->short_ref_count; + dst->prev_poc_msb = src->prev_poc_msb; + dst->prev_poc_lsb = src->prev_poc_lsb; + dst->prev_frame_num_offset = src->prev_frame_num_offset; + dst->prev_frame_num = src->prev_frame_num; + dst->short_ref_count = src->short_ref_count; memcpy(dst->short_ref, src->short_ref, sizeof(dst->short_ref)); memcpy(dst->long_ref, src->long_ref, sizeof(dst->long_ref)); @@ -2608,16 +2836,16 @@ int ff_h264_get_profile(SPS *sps) { int profile = sps->profile_idc; - switch(sps->profile_idc) { + switch (sps->profile_idc) { case FF_PROFILE_H264_BASELINE: // constraint_set1_flag set to 1 - profile |= (sps->constraint_set_flags & 1<<1) ? FF_PROFILE_H264_CONSTRAINED : 0; + profile |= (sps->constraint_set_flags & 1 << 1) ? FF_PROFILE_H264_CONSTRAINED : 0; break; case FF_PROFILE_H264_HIGH_10: case FF_PROFILE_H264_HIGH_422: case FF_PROFILE_H264_HIGH_444_PREDICTIVE: // constraint_set3_flag set to 1 - profile |= (sps->constraint_set_flags & 1<<3) ? FF_PROFILE_H264_INTRA : 0; + profile |= (sps->constraint_set_flags & 1 << 3) ? FF_PROFILE_H264_INTRA : 0; break; } @@ -2629,13 +2857,15 @@ int ff_h264_get_profile(SPS *sps) * This will also call ff_MPV_common_init() and frame_start() as needed. * * @param h h264context - * @param h0 h264 master context (differs from 'h' when doing sliced based parallel decoding) + * @param h0 h264 master context (differs from 'h' when doing sliced based + * parallel decoding) * * @return 0 if okay, <0 if an error occurred, 1 if decoding must not be multithreaded */ -static int decode_slice_header(H264Context *h, H264Context *h0){ - MpegEncContext * const s = &h->s; - MpegEncContext * const s0 = &h0->s; +static int decode_slice_header(H264Context *h, H264Context *h0) +{ + MpegEncContext *const s = &h->s; + MpegEncContext *const s0 = &h0->s; unsigned int first_mb_in_slice; unsigned int pps_id; int num_ref_idx_active_override_flag; @@ -2643,63 +2873,71 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ int default_ref_list_done = 0; int last_pic_structure; - s->dropable= h->nal_ref_idc == 0; + s->dropable = h->nal_ref_idc == 0; /* FIXME: 2tap qpel isn't implemented for high bit depth. */ - if((s->avctx->flags2 & CODEC_FLAG2_FAST) && !h->nal_ref_idc && !h->pixel_shift){ - s->me.qpel_put= s->dsp.put_2tap_qpel_pixels_tab; - s->me.qpel_avg= s->dsp.avg_2tap_qpel_pixels_tab; - }else{ - s->me.qpel_put= s->dsp.put_h264_qpel_pixels_tab; - s->me.qpel_avg= s->dsp.avg_h264_qpel_pixels_tab; + if ((s->avctx->flags2 & CODEC_FLAG2_FAST) && + !h->nal_ref_idc && !h->pixel_shift) { + s->me.qpel_put = s->dsp.put_2tap_qpel_pixels_tab; + s->me.qpel_avg = s->dsp.avg_2tap_qpel_pixels_tab; + } else { + s->me.qpel_put = s->dsp.put_h264_qpel_pixels_tab; + 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(&s->gb); - if(first_mb_in_slice == 0){ //FIXME better field boundary detection - if(h0->current_slice && FIELD_PICTURE){ + if (first_mb_in_slice == 0) { // FIXME better field boundary detection + if (h0->current_slice && FIELD_PICTURE) { field_end(h, 1); } h0->current_slice = 0; if (!s0->first_field) - s->current_picture_ptr= NULL; + s->current_picture_ptr = NULL; } - slice_type= get_ue_golomb_31(&s->gb); - if(slice_type > 9){ - av_log(h->s.avctx, AV_LOG_ERROR, "slice type too large (%d) at %d %d\n", h->slice_type, s->mb_x, s->mb_y); + slice_type = get_ue_golomb_31(&s->gb); + if (slice_type > 9) { + av_log(h->s.avctx, AV_LOG_ERROR, + "slice type too large (%d) at %d %d\n", + h->slice_type, s->mb_x, s->mb_y); return -1; } - if(slice_type > 4){ + if (slice_type > 4) { slice_type -= 5; - h->slice_type_fixed=1; - }else - h->slice_type_fixed=0; + h->slice_type_fixed = 1; + } else + h->slice_type_fixed = 0; - slice_type= golomb_to_pict_type[ slice_type ]; - if (slice_type == AV_PICTURE_TYPE_I - || (h0->current_slice != 0 && slice_type == h0->last_slice_type) ) { + slice_type = golomb_to_pict_type[slice_type]; + if (slice_type == AV_PICTURE_TYPE_I || + (h0->current_slice != 0 && slice_type == h0->last_slice_type)) { default_ref_list_done = 1; } - h->slice_type= slice_type; - h->slice_type_nos= slice_type & 3; + h->slice_type = slice_type; + h->slice_type_nos = slice_type & 3; - s->pict_type= h->slice_type; // to make a few old functions happy, it's wrong though + // to make a few old functions happy, it's wrong though + s->pict_type = h->slice_type; - pps_id= get_ue_golomb(&s->gb); - if(pps_id>=MAX_PPS_COUNT){ + pps_id = get_ue_golomb(&s->gb); + if (pps_id >= MAX_PPS_COUNT) { av_log(h->s.avctx, AV_LOG_ERROR, "pps_id out of range\n"); return -1; } - if(!h0->pps_buffers[pps_id]) { - av_log(h->s.avctx, AV_LOG_ERROR, "non-existing PPS %u referenced\n", pps_id); + if (!h0->pps_buffers[pps_id]) { + av_log(h->s.avctx, AV_LOG_ERROR, + "non-existing PPS %u referenced\n", + pps_id); return -1; } - h->pps= *h0->pps_buffers[pps_id]; + h->pps = *h0->pps_buffers[pps_id]; - if(!h0->sps_buffers[h->pps.sps_id]) { - av_log(h->s.avctx, AV_LOG_ERROR, "non-existing SPS %u referenced\n", h->pps.sps_id); + if (!h0->sps_buffers[h->pps.sps_id]) { + av_log(h->s.avctx, AV_LOG_ERROR, + "non-existing SPS %u referenced\n", + h->pps.sps_id); return -1; } h->sps = *h0->sps_buffers[h->pps.sps_id]; @@ -2708,24 +2946,25 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ s->avctx->level = h->sps.level_idc; s->avctx->refs = h->sps.ref_frame_count; - s->mb_width= h->sps.mb_width; - s->mb_height= h->sps.mb_height * (2 - h->sps.frame_mbs_only_flag); + s->mb_width = h->sps.mb_width; + s->mb_height = h->sps.mb_height * (2 - h->sps.frame_mbs_only_flag); - h->b_stride= s->mb_width*4; + h->b_stride = s->mb_width * 4; 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<sps.frame_mbs_only_flag) - s->height= 16*s->mb_height - (1<chroma_y_shift)*FFMIN(h->sps.crop_bottom, (16>>s->chroma_y_shift)-1); + 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<chroma_y_shift)*FFMIN(h->sps.crop_bottom, (16>>s->chroma_y_shift)-1); + s->height = 16 * s->mb_height - (2 << s->chroma_y_shift) * FFMIN(h->sps.crop_bottom, (16 >> s->chroma_y_shift) - 1); - if (s->context_initialized - && ( s->width != s->avctx->width || s->height != s->avctx->height - || 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); + if (s->context_initialized && + (s->width != s->avctx->width || s->height != s->avctx->height || + 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); return -1; // width / height changed during parallelized decoding } free_tables(h, 0); @@ -2734,78 +2973,83 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ } if (!s->context_initialized) { if (h != h0) { - av_log(h->s.avctx, AV_LOG_ERROR, "Cannot (re-)initialize context during parallel decoding.\n"); + 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->sample_aspect_ratio= h->sps.sar; + s->avctx->sample_aspect_ratio = h->sps.sar; av_assert0(s->avctx->sample_aspect_ratio.den); - if(h->sps.video_signal_type_present_flag){ - s->avctx->color_range = h->sps.full_range ? AVCOL_RANGE_JPEG : AVCOL_RANGE_MPEG; - if(h->sps.colour_description_present_flag){ + if (h->sps.video_signal_type_present_flag) { + s->avctx->color_range = h->sps.full_range ? 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; s->avctx->colorspace = h->sps.colorspace; } } - if(h->sps.timing_info_present_flag){ - int64_t den= h->sps.time_scale; - if(h->x264_build < 44U) + if (h->sps.timing_info_present_flag) { + int64_t den = h->sps.time_scale; + if (h->x264_build < 44U) den *= 2; av_reduce(&s->avctx->time_base.num, &s->avctx->time_base.den, - h->sps.num_units_in_tick, den, 1<<30); + h->sps.num_units_in_tick, den, 1 << 30); } switch (h->sps.bit_depth_luma) { - case 9 : - 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) { - 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; - case 8: - if (CHROMA444){ - 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; - } else if (CHROMA422) { - s->avctx->pix_fmt = s->avctx->color_range == AVCOL_RANGE_JPEG ? PIX_FMT_YUVJ422P : PIX_FMT_YUV422P; - }else{ - s->avctx->pix_fmt = s->avctx->get_format(s->avctx, - s->avctx->codec->pix_fmts ? - s->avctx->codec->pix_fmts : - s->avctx->color_range == AVCOL_RANGE_JPEG ? - hwaccel_pixfmt_list_h264_jpeg_420 : - ff_hwaccel_pixfmt_list_420); - } - break; - default: - av_log(s->avctx, AV_LOG_ERROR, - "Unsupported bit depth: %d\n", h->sps.bit_depth_luma); - return AVERROR_INVALIDDATA; + case 9: + 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) { + 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; + case 8: + if (CHROMA444) { + 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; + } else if (CHROMA422) { + s->avctx->pix_fmt = s->avctx->color_range == AVCOL_RANGE_JPEG ? PIX_FMT_YUVJ422P + : PIX_FMT_YUV422P; + } else { + s->avctx->pix_fmt = s->avctx->get_format(s->avctx, + s->avctx->codec->pix_fmts ? + s->avctx->codec->pix_fmts : + s->avctx->color_range == AVCOL_RANGE_JPEG ? + hwaccel_pixfmt_list_h264_jpeg_420 : + ff_hwaccel_pixfmt_list_420); + } + break; + default: + av_log(s->avctx, AV_LOG_ERROR, + "Unsupported bit depth: %d\n", h->sps.bit_depth_luma); + return AVERROR_INVALIDDATA; } - s->avctx->hwaccel = ff_find_hwaccel(s->avctx->codec->id, s->avctx->pix_fmt); + s->avctx->hwaccel = ff_find_hwaccel(s->avctx->codec->id, + s->avctx->pix_fmt); if (ff_MPV_common_init(s) < 0) { av_log(h->s.avctx, AV_LOG_ERROR, "ff_MPV_common_init() failed.\n"); @@ -2816,65 +3060,70 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ init_scan_tables(h); if (ff_h264_alloc_tables(h) < 0) { - av_log(h->s.avctx, AV_LOG_ERROR, "Could not allocate memory for h264\n"); + av_log(h->s.avctx, AV_LOG_ERROR, + "Could not allocate memory for h264\n"); return AVERROR(ENOMEM); } - if (!HAVE_THREADS || !(s->avctx->active_thread_type&FF_THREAD_SLICE)) { + if (!HAVE_THREADS || !(s->avctx->active_thread_type & FF_THREAD_SLICE)) { if (context_init(h) < 0) { av_log(h->s.avctx, AV_LOG_ERROR, "context_init() failed.\n"); return -1; } } else { - for(i = 1; i < s->slice_context_count; i++) { + for (i = 1; i < s->slice_context_count; i++) { H264Context *c; c = h->thread_context[i] = av_malloc(sizeof(H264Context)); memcpy(c, h->s.thread_context[i], sizeof(MpegEncContext)); memset(&c->s + 1, 0, sizeof(H264Context) - sizeof(MpegEncContext)); - c->h264dsp = h->h264dsp; - c->sps = h->sps; - c->pps = h->pps; + c->h264dsp = h->h264dsp; + c->sps = h->sps; + c->pps = h->pps; c->pixel_shift = h->pixel_shift; init_scan_tables(c); clone_tables(c, h, i); } - for(i = 0; i < s->slice_context_count; i++) + for (i = 0; i < s->slice_context_count; i++) if (context_init(h->thread_context[i]) < 0) { - av_log(h->s.avctx, AV_LOG_ERROR, "context_init() failed.\n"); + av_log(h->s.avctx, AV_LOG_ERROR, + "context_init() failed.\n"); return -1; } } } - if(h == h0 && h->dequant_coeff_pps != pps_id){ + if (h == h0 && h->dequant_coeff_pps != pps_id) { h->dequant_coeff_pps = pps_id; init_dequant_tables(h); } - h->frame_num= get_bits(&s->gb, h->sps.log2_max_frame_num); + h->frame_num = get_bits(&s->gb, h->sps.log2_max_frame_num); - h->mb_mbaff = 0; - h->mb_aff_frame = 0; + h->mb_mbaff = 0; + h->mb_aff_frame = 0; last_pic_structure = s0->picture_structure; - if(h->sps.frame_mbs_only_flag){ - s->picture_structure= PICT_FRAME; - }else{ - if(get_bits1(&s->gb)) { //field_pic_flag - s->picture_structure= PICT_TOP_FIELD + get_bits1(&s->gb); //bottom_field_flag + if (h->sps.frame_mbs_only_flag) { + s->picture_structure = PICT_FRAME; + } else { + if (get_bits1(&s->gb)) { // field_pic_flag + s->picture_structure = PICT_TOP_FIELD + get_bits1(&s->gb); // bottom_field_flag } else { - s->picture_structure= PICT_FRAME; - h->mb_aff_frame = h->sps.mb_aff; + s->picture_structure = PICT_FRAME; + h->mb_aff_frame = h->sps.mb_aff; } } - h->mb_field_decoding_flag= s->picture_structure != PICT_FRAME; + h->mb_field_decoding_flag = s->picture_structure != PICT_FRAME; - 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) { - int unwrap_prev_frame_num = h->prev_frame_num, max_frame_num = 1<sps.log2_max_frame_num; + 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) { + int unwrap_prev_frame_num = h->prev_frame_num; + int 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; + if (unwrap_prev_frame_num > h->frame_num) + unwrap_prev_frame_num -= max_frame_num; if ((h->frame_num - unwrap_prev_frame_num) > h->sps.ref_frame_count) { unwrap_prev_frame_num = (h->frame_num - h->sps.ref_frame_count) - 1; @@ -2885,15 +3134,16 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ } } - while(h->frame_num != h->prev_frame_num && - h->frame_num != (h->prev_frame_num+1)%(1<sps.log2_max_frame_num)){ + while (h->frame_num != h->prev_frame_num && + 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); + av_log(h->s.avctx, AV_LOG_DEBUG, "Frame num gap %d %d\n", + h->frame_num, h->prev_frame_num); if (ff_h264_frame_start(h) < 0) return -1; h->prev_frame_num++; - h->prev_frame_num %= 1<sps.log2_max_frame_num; - s->current_picture_ptr->frame_num= h->prev_frame_num; + h->prev_frame_num %= 1 << h->sps.log2_max_frame_num; + s->current_picture_ptr->frame_num = h->prev_frame_num; ff_thread_report_progress(&s->current_picture_ptr->f, INT_MAX, 0); ff_thread_report_progress(&s->current_picture_ptr->f, INT_MAX, 1); ff_generate_sliding_window_mmcos(h); @@ -2909,9 +3159,9 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ if (h->short_ref_count) { if (prev) { av_image_copy(h->short_ref[0]->f.data, h->short_ref[0]->f.linesize, - (const uint8_t**)prev->f.data, prev->f.linesize, - s->avctx->pix_fmt, s->mb_width*16, s->mb_height*16); - h->short_ref[0]->poc = prev->poc+2; + (const uint8_t **)prev->f.data, prev->f.linesize, + s->avctx->pix_fmt, s->mb_width * 16, s->mb_height * 16); + h->short_ref[0]->poc = prev->poc + 2; } h->short_ref[0]->frame_num = h->prev_frame_num; } @@ -2925,39 +3175,31 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ /* figure out if we have a complementary field pair */ if (!FIELD_PICTURE || s->picture_structure == last_pic_structure) { - /* - * Previous field is unmatched. Don't display it, but let it - * remain for reference if marked as such. - */ + /* Previous field is unmatched. Don't display it, but let it + * remain for reference if marked as such. */ s0->current_picture_ptr = NULL; - s0->first_field = FIELD_PICTURE; - + 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) { - /* - * This and previous field were reference, but had - * different frame_nums. Consider this field first in - * pair. Throw away previous field except for reference - * purposes. - */ - s0->first_field = 1; + s0->current_picture_ptr->f.reference && + s0->current_picture_ptr->frame_num != h->frame_num) { + /* This and the previous field were reference, but had + * different frame_nums. Consider this field first in pair. + * Throw away previous one except for reference purposes. */ + s0->first_field = 1; s0->current_picture_ptr = NULL; - } else { /* Second field in complementary pair */ s0->first_field = 0; } } - } else { /* Frame or first field in a potentially complementary pair */ assert(!s0->current_picture_ptr); s0->first_field = FIELD_PICTURE; } - if(!FIELD_PICTURE || s0->first_field) { + if (!FIELD_PICTURE || s0->first_field) { if (ff_h264_frame_start(h) < 0) { s0->first_field = 0; return -1; @@ -2966,72 +3208,68 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ ff_release_unused_pictures(s, 0); } } - if(h != h0) + if (h != h0) clone_slice(h, h0); - s->current_picture_ptr->frame_num= h->frame_num; //FIXME frame_num cleanup + s->current_picture_ptr->frame_num = h->frame_num; // FIXME frame_num cleanup assert(s->mb_num == s->mb_width * s->mb_height); - if(first_mb_in_slice << FIELD_OR_MBAFF_PICTURE >= s->mb_num || - first_mb_in_slice >= s->mb_num){ + if (first_mb_in_slice << FIELD_OR_MBAFF_PICTURE >= s->mb_num || + first_mb_in_slice >= s->mb_num) { av_log(h->s.avctx, AV_LOG_ERROR, "first_mb_in_slice overflow\n"); return -1; } - s->resync_mb_x = s->mb_x = first_mb_in_slice % s->mb_width; + s->resync_mb_x = s->mb_x = first_mb_in_slice % s->mb_width; s->resync_mb_y = s->mb_y = (first_mb_in_slice / s->mb_width) << FIELD_OR_MBAFF_PICTURE; if (s->picture_structure == PICT_BOTTOM_FIELD) s->resync_mb_y = s->mb_y = s->mb_y + 1; assert(s->mb_y < s->mb_height); - if(s->picture_structure==PICT_FRAME){ - h->curr_pic_num= h->frame_num; - h->max_pic_num= 1<< h->sps.log2_max_frame_num; - }else{ - h->curr_pic_num= 2*h->frame_num + 1; - h->max_pic_num= 1<<(h->sps.log2_max_frame_num + 1); + if (s->picture_structure == PICT_FRAME) { + h->curr_pic_num = h->frame_num; + h->max_pic_num = 1 << h->sps.log2_max_frame_num; + } else { + h->curr_pic_num = 2 * h->frame_num + 1; + h->max_pic_num = 1 << (h->sps.log2_max_frame_num + 1); } - if(h->nal_unit_type == NAL_IDR_SLICE){ + if (h->nal_unit_type == NAL_IDR_SLICE) get_ue_golomb(&s->gb); /* idr_pic_id */ + + if (h->sps.poc_type == 0) { + h->poc_lsb = get_bits(&s->gb, h->sps.log2_max_poc_lsb); + + if (h->pps.pic_order_present == 1 && s->picture_structure == PICT_FRAME) + h->delta_poc_bottom = get_se_golomb(&s->gb); } - if(h->sps.poc_type==0){ - h->poc_lsb= get_bits(&s->gb, h->sps.log2_max_poc_lsb); + if (h->sps.poc_type == 1 && !h->sps.delta_pic_order_always_zero_flag) { + h->delta_poc[0] = get_se_golomb(&s->gb); - if(h->pps.pic_order_present==1 && s->picture_structure==PICT_FRAME){ - h->delta_poc_bottom= get_se_golomb(&s->gb); - } - } - - if(h->sps.poc_type==1 && !h->sps.delta_pic_order_always_zero_flag){ - h->delta_poc[0]= get_se_golomb(&s->gb); - - if(h->pps.pic_order_present==1 && s->picture_structure==PICT_FRAME) - h->delta_poc[1]= get_se_golomb(&s->gb); + if (h->pps.pic_order_present == 1 && s->picture_structure == PICT_FRAME) + h->delta_poc[1] = get_se_golomb(&s->gb); } init_poc(h); - if(h->pps.redundant_pic_cnt_present){ - h->redundant_pic_count= get_ue_golomb(&s->gb); - } + if (h->pps.redundant_pic_cnt_present) + h->redundant_pic_count = get_ue_golomb(&s->gb); - //set defaults, might be overridden a few lines later - h->ref_count[0]= h->pps.ref_count[0]; - h->ref_count[1]= h->pps.ref_count[1]; + // set defaults, might be overridden a few lines later + h->ref_count[0] = h->pps.ref_count[0]; + h->ref_count[1] = h->pps.ref_count[1]; - if(h->slice_type_nos != AV_PICTURE_TYPE_I){ + if (h->slice_type_nos != AV_PICTURE_TYPE_I) { int max_refs = s->picture_structure == PICT_FRAME ? 16 : 32; - if(h->slice_type_nos == AV_PICTURE_TYPE_B){ - h->direct_spatial_mv_pred= get_bits1(&s->gb); - } - num_ref_idx_active_override_flag= get_bits1(&s->gb); + if (h->slice_type_nos == AV_PICTURE_TYPE_B) + h->direct_spatial_mv_pred = get_bits1(&s->gb); + num_ref_idx_active_override_flag = get_bits1(&s->gb); - if(num_ref_idx_active_override_flag){ - h->ref_count[0]= get_ue_golomb(&s->gb) + 1; - if(h->slice_type_nos==AV_PICTURE_TYPE_B) - h->ref_count[1]= get_ue_golomb(&s->gb) + 1; + if (num_ref_idx_active_override_flag) { + h->ref_count[0] = get_ue_golomb(&s->gb) + 1; + if (h->slice_type_nos == AV_PICTURE_TYPE_B) + h->ref_count[1] = get_ue_golomb(&s->gb) + 1; } if (h->ref_count[0] > max_refs || h->ref_count[1] > max_refs) { @@ -3040,37 +3278,39 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ return AVERROR_INVALIDDATA; } - if(h->slice_type_nos == AV_PICTURE_TYPE_B) - h->list_count= 2; + if (h->slice_type_nos == AV_PICTURE_TYPE_B) + h->list_count = 2; else - h->list_count= 1; - }else - h->list_count= 0; + h->list_count = 1; + } else + h->list_count = 0; - if(!default_ref_list_done){ + if (!default_ref_list_done) ff_h264_fill_default_ref_list(h); - } - if(h->slice_type_nos!=AV_PICTURE_TYPE_I && ff_h264_decode_ref_pic_list_reordering(h) < 0) { - h->ref_count[1]= h->ref_count[0]= 0; + if (h->slice_type_nos != AV_PICTURE_TYPE_I && + ff_h264_decode_ref_pic_list_reordering(h) < 0) { + h->ref_count[1] = h->ref_count[0] = 0; return -1; } - if(h->slice_type_nos!=AV_PICTURE_TYPE_I){ - s->last_picture_ptr= &h->ref_list[0][0]; + if (h->slice_type_nos != AV_PICTURE_TYPE_I) { + s->last_picture_ptr = &h->ref_list[0][0]; ff_copy_picture(&s->last_picture, s->last_picture_ptr); } - if(h->slice_type_nos==AV_PICTURE_TYPE_B){ - s->next_picture_ptr= &h->ref_list[1][0]; + if (h->slice_type_nos == AV_PICTURE_TYPE_B) { + s->next_picture_ptr = &h->ref_list[1][0]; ff_copy_picture(&s->next_picture, s->next_picture_ptr); } - if( (h->pps.weighted_pred && h->slice_type_nos == AV_PICTURE_TYPE_P ) - || (h->pps.weighted_bipred_idc==1 && h->slice_type_nos== AV_PICTURE_TYPE_B ) ) + if ((h->pps.weighted_pred && h->slice_type_nos == AV_PICTURE_TYPE_P) || + (h->pps.weighted_bipred_idc == 1 && + h->slice_type_nos == AV_PICTURE_TYPE_B)) pred_weight_table(h); - else if(h->pps.weighted_bipred_idc==2 && h->slice_type_nos== AV_PICTURE_TYPE_B){ + else if (h->pps.weighted_bipred_idc == 2 && + h->slice_type_nos == AV_PICTURE_TYPE_B) { implicit_weight_table(h, -1); - }else { + } else { h->use_weight = 0; for (i = 0; i < 2; i++) { h->luma_weight_flag[i] = 0; @@ -3078,163 +3318,176 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ } } - if(h->nal_ref_idc && ff_h264_decode_ref_pic_marking(h0, &s->gb) < 0 && - (s->avctx->err_recognition & AV_EF_EXPLODE)) + if (h->nal_ref_idc && ff_h264_decode_ref_pic_marking(h0, &s->gb) < 0 && + (s->avctx->err_recognition & AV_EF_EXPLODE)) return AVERROR_INVALIDDATA; - if(FRAME_MBAFF){ + if (FRAME_MBAFF) { ff_h264_fill_mbaff_ref_list(h); - if(h->pps.weighted_bipred_idc==2 && h->slice_type_nos== AV_PICTURE_TYPE_B){ + if (h->pps.weighted_bipred_idc == 2 && h->slice_type_nos == AV_PICTURE_TYPE_B) { implicit_weight_table(h, 0); implicit_weight_table(h, 1); } } - if(h->slice_type_nos==AV_PICTURE_TYPE_B && !h->direct_spatial_mv_pred) + if (h->slice_type_nos == AV_PICTURE_TYPE_B && !h->direct_spatial_mv_pred) ff_h264_direct_dist_scale_factor(h); ff_h264_direct_ref_list_init(h); - if( h->slice_type_nos != AV_PICTURE_TYPE_I && h->pps.cabac ){ + if (h->slice_type_nos != AV_PICTURE_TYPE_I && h->pps.cabac) { tmp = get_ue_golomb_31(&s->gb); - if(tmp > 2){ + if (tmp > 2) { av_log(s->avctx, AV_LOG_ERROR, "cabac_init_idc overflow\n"); return -1; } - h->cabac_init_idc= tmp; + h->cabac_init_idc = tmp; } h->last_qscale_diff = 0; tmp = h->pps.init_qp + get_se_golomb(&s->gb); - if(tmp>51+6*(h->sps.bit_depth_luma-8)){ + if (tmp > 51 + 6 * (h->sps.bit_depth_luma - 8)) { av_log(s->avctx, AV_LOG_ERROR, "QP %u out of range\n", tmp); return -1; } - s->qscale= tmp; + s->qscale = tmp; h->chroma_qp[0] = get_chroma_qp(h, 0, s->qscale); h->chroma_qp[1] = get_chroma_qp(h, 1, s->qscale); - //FIXME qscale / qp ... stuff - if(h->slice_type == AV_PICTURE_TYPE_SP){ + // FIXME qscale / qp ... stuff + if (h->slice_type == AV_PICTURE_TYPE_SP) get_bits1(&s->gb); /* sp_for_switch_flag */ - } - if(h->slice_type==AV_PICTURE_TYPE_SP || h->slice_type == AV_PICTURE_TYPE_SI){ + if (h->slice_type == AV_PICTURE_TYPE_SP || + h->slice_type == AV_PICTURE_TYPE_SI) get_se_golomb(&s->gb); /* slice_qs_delta */ - } - h->deblocking_filter = 1; + h->deblocking_filter = 1; h->slice_alpha_c0_offset = 52; - h->slice_beta_offset = 52; - if( h->pps.deblocking_filter_parameters_present ) { - tmp= get_ue_golomb_31(&s->gb); - if(tmp > 2){ - av_log(s->avctx, AV_LOG_ERROR, "deblocking_filter_idc %u out of range\n", tmp); + h->slice_beta_offset = 52; + if (h->pps.deblocking_filter_parameters_present) { + tmp = get_ue_golomb_31(&s->gb); + if (tmp > 2) { + av_log(s->avctx, AV_LOG_ERROR, + "deblocking_filter_idc %u out of range\n", tmp); return -1; } - h->deblocking_filter= tmp; - if(h->deblocking_filter < 2) - h->deblocking_filter^= 1; // 1<->0 + h->deblocking_filter = tmp; + if (h->deblocking_filter < 2) + h->deblocking_filter ^= 1; // 1<->0 - if( h->deblocking_filter ) { + if (h->deblocking_filter) { h->slice_alpha_c0_offset += get_se_golomb(&s->gb) << 1; h->slice_beta_offset += get_se_golomb(&s->gb) << 1; - if( h->slice_alpha_c0_offset > 104U - || h->slice_beta_offset > 104U){ - av_log(s->avctx, AV_LOG_ERROR, "deblocking filter parameters %d %d out of range\n", h->slice_alpha_c0_offset, h->slice_beta_offset); + if (h->slice_alpha_c0_offset > 104U || + h->slice_beta_offset > 104U) { + av_log(s->avctx, AV_LOG_ERROR, + "deblocking filter parameters %d %d out of range\n", + h->slice_alpha_c0_offset, h->slice_beta_offset); return -1; } } } - if( s->avctx->skip_loop_filter >= AVDISCARD_ALL - ||(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY && h->slice_type_nos != AV_PICTURE_TYPE_I) - ||(s->avctx->skip_loop_filter >= AVDISCARD_BIDIR && h->slice_type_nos == AV_PICTURE_TYPE_B) - ||(s->avctx->skip_loop_filter >= AVDISCARD_NONREF && h->nal_ref_idc == 0)) - h->deblocking_filter= 0; + if (s->avctx->skip_loop_filter >= AVDISCARD_ALL || + (s->avctx->skip_loop_filter >= AVDISCARD_NONKEY && + h->slice_type_nos != AV_PICTURE_TYPE_I) || + (s->avctx->skip_loop_filter >= AVDISCARD_BIDIR && + h->slice_type_nos == AV_PICTURE_TYPE_B) || + (s->avctx->skip_loop_filter >= AVDISCARD_NONREF && + h->nal_ref_idc == 0)) + h->deblocking_filter = 0; - if(h->deblocking_filter == 1 && h0->max_contexts > 1) { - if(s->avctx->flags2 & CODEC_FLAG2_FAST) { + if (h->deblocking_filter == 1 && h0->max_contexts > 1) { + if (s->avctx->flags2 & CODEC_FLAG2_FAST) { /* Cheat slightly for speed: - Do not bother to deblock across slices. */ + * Do not bother to deblock across slices. */ h->deblocking_filter = 2; } else { h0->max_contexts = 1; - if(!h0->single_decode_warning) { - av_log(s->avctx, AV_LOG_INFO, "Cannot parallelize deblocking type 1, decoding such frames in sequential order\n"); + if (!h0->single_decode_warning) { + av_log(s->avctx, AV_LOG_INFO, + "Cannot parallelize deblocking type 1, decoding such frames in sequential order\n"); h0->single_decode_warning = 1; } if (h != h0) { - av_log(h->s.avctx, AV_LOG_ERROR, "Deblocking switched inside frame.\n"); + av_log(h->s.avctx, AV_LOG_ERROR, + "Deblocking switched inside frame.\n"); return 1; } } } - h->qp_thresh = 15 + 52 - FFMIN(h->slice_alpha_c0_offset, h->slice_beta_offset) - - FFMAX3(0, h->pps.chroma_qp_index_offset[0], h->pps.chroma_qp_index_offset[1]) - + 6 * (h->sps.bit_depth_luma - 8); - -#if 0 //FMO - if( h->pps.num_slice_groups > 1 && h->pps.mb_slice_group_map_type >= 3 && h->pps.mb_slice_group_map_type <= 5) - slice_group_change_cycle= get_bits(&s->gb, ?); -#endif + h->qp_thresh = 15 + 52 - + FFMIN(h->slice_alpha_c0_offset, h->slice_beta_offset) - + FFMAX3(0, + h->pps.chroma_qp_index_offset[0], + h->pps.chroma_qp_index_offset[1]) + + 6 * (h->sps.bit_depth_luma - 8); 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 >= MAX_SLICES) { + av_log(s->avctx, AV_LOG_ERROR, + "Too many slices, increase MAX_SLICES and recompile\n"); } - for(j=0; j<2; j++){ + for (j = 0; j < 2; j++) { int id_list[16]; - int *ref2frm= h->ref2frm[h->slice_num&(MAX_SLICES-1)][j]; - for(i=0; i<16; i++){ - id_list[i]= 60; + int *ref2frm = h->ref2frm[h->slice_num & (MAX_SLICES - 1)][j]; + for (i = 0; i < 16; i++) { + id_list[i] = 60; if (h->ref_list[j][i].f.data[0]) { int k; uint8_t *base = h->ref_list[j][i].f.base[0]; - for(k=0; kshort_ref_count; k++) + for (k = 0; k < h->short_ref_count; k++) if (h->short_ref[k]->f.base[0] == base) { - id_list[i]= k; + id_list[i] = k; break; } - for(k=0; klong_ref_count; k++) + for (k = 0; k < h->long_ref_count; k++) if (h->long_ref[k] && h->long_ref[k]->f.base[0] == base) { - id_list[i]= h->short_ref_count + k; + id_list[i] = h->short_ref_count + k; break; } } } - ref2frm[0]= - ref2frm[1]= -1; - for(i=0; i<16; i++) - ref2frm[i+2]= 4*id_list[i] - + (h->ref_list[j][i].f.reference & 3); - ref2frm[18+0]= - ref2frm[18+1]= -1; - for(i=16; i<48; i++) - ref2frm[i+4]= 4*id_list[(i-16)>>1] - + (h->ref_list[j][i].f.reference & 3); + ref2frm[0] = + ref2frm[1] = -1; + for (i = 0; i < 16; i++) + ref2frm[i + 2] = 4 * id_list[i] + + (h->ref_list[j][i].f.reference & 3); + ref2frm[18 + 0] = + ref2frm[18 + 1] = -1; + for (i = 16; i < 48; i++) + ref2frm[i + 4] = 4 * id_list[(i - 16) >> 1] + + (h->ref_list[j][i].f.reference & 3); } - //FIXME: fix draw_edges+PAFF+frame threads - h->emu_edge_width= (s->flags&CODEC_FLAG_EMU_EDGE || (!h->sps.frame_mbs_only_flag && s->avctx->active_thread_type)) ? 0 : 16; - h->emu_edge_height= (FRAME_MBAFF || FIELD_PICTURE) ? 0 : h->emu_edge_width; + // FIXME: fix draw_edges + PAFF + frame threads + h->emu_edge_width = (s->flags & CODEC_FLAG_EMU_EDGE || + (!h->sps.frame_mbs_only_flag && + s->avctx->active_thread_type)) + ? 0 : 16; + h->emu_edge_height = (FRAME_MBAFF || FIELD_PICTURE) ? 0 : h->emu_edge_width; - if(s->avctx->debug&FF_DEBUG_PICT_INFO){ - av_log(h->s.avctx, AV_LOG_DEBUG, "slice:%d %s mb:%d %c%s%s pps:%u frame:%d poc:%d/%d ref:%d/%d qp:%d loop:%d:%d:%d weight:%d%s %s\n", + if (s->avctx->debug & FF_DEBUG_PICT_INFO) { + av_log(h->s.avctx, AV_LOG_DEBUG, + "slice:%d %s mb:%d %c%s%s pps:%u frame:%d poc:%d/%d ref:%d/%d qp:%d loop:%d:%d:%d weight:%d%s %s\n", h->slice_num, - (s->picture_structure==PICT_FRAME ? "F" : s->picture_structure==PICT_TOP_FIELD ? "T" : "B"), + (s->picture_structure == PICT_FRAME ? "F" : s->picture_structure == PICT_TOP_FIELD ? "T" : "B"), first_mb_in_slice, - av_get_picture_type_char(h->slice_type), h->slice_type_fixed ? " fix" : "", h->nal_unit_type == NAL_IDR_SLICE ? " IDR" : "", + av_get_picture_type_char(h->slice_type), + h->slice_type_fixed ? " fix" : "", + h->nal_unit_type == NAL_IDR_SLICE ? " IDR" : "", pps_id, h->frame_num, - s->current_picture_ptr->field_poc[0], s->current_picture_ptr->field_poc[1], + s->current_picture_ptr->field_poc[0], + s->current_picture_ptr->field_poc[1], h->ref_count[0], h->ref_count[1], s->qscale, - h->deblocking_filter, h->slice_alpha_c0_offset/2-26, h->slice_beta_offset/2-26, + h->deblocking_filter, + h->slice_alpha_c0_offset / 2 - 26, h->slice_beta_offset / 2 - 26, h->use_weight, - h->use_weight==1 && h->use_weight_chroma ? "c" : "", - h->slice_type == AV_PICTURE_TYPE_B ? (h->direct_spatial_mv_pred ? "SPAT" : "TEMP") : "" - ); + h->use_weight == 1 && h->use_weight_chroma ? "c" : "", + h->slice_type == AV_PICTURE_TYPE_B ? (h->direct_spatial_mv_pred ? "SPAT" : "TEMP") : ""); } return 0; @@ -3243,88 +3496,99 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ int ff_h264_get_slice_type(const H264Context *h) { switch (h->slice_type) { - case AV_PICTURE_TYPE_P: return 0; - case AV_PICTURE_TYPE_B: return 1; - case AV_PICTURE_TYPE_I: return 2; - case AV_PICTURE_TYPE_SP: return 3; - case AV_PICTURE_TYPE_SI: return 4; - default: return -1; + case AV_PICTURE_TYPE_P: + return 0; + case AV_PICTURE_TYPE_B: + return 1; + case AV_PICTURE_TYPE_I: + return 2; + case AV_PICTURE_TYPE_SP: + return 3; + case AV_PICTURE_TYPE_SI: + return 4; + default: + return -1; } } -static av_always_inline void fill_filter_caches_inter(H264Context *h, MpegEncContext * const s, int mb_type, int top_xy, - int left_xy[LEFT_MBS], int top_type, int left_type[LEFT_MBS], int mb_xy, int list) +static av_always_inline void fill_filter_caches_inter(H264Context *h, + MpegEncContext *const s, + int mb_type, int top_xy, + int left_xy[LEFT_MBS], + int top_type, + int left_type[LEFT_MBS], + int mb_xy, int list) { int b_stride = h->b_stride; - int16_t (*mv_dst)[2] = &h->mv_cache[list][scan8[0]]; + int16_t(*mv_dst)[2] = &h->mv_cache[list][scan8[0]]; int8_t *ref_cache = &h->ref_cache[list][scan8[0]]; - if(IS_INTER(mb_type) || IS_DIRECT(mb_type)){ - if(USES_LIST(top_type, list)){ - const int b_xy= h->mb2b_xy[top_xy] + 3*b_stride; - const int b8_xy= 4*top_xy + 2; - int (*ref2frm)[64] = h->ref2frm[ h->slice_table[top_xy]&(MAX_SLICES-1) ][0] + (MB_MBAFF ? 20 : 2); - AV_COPY128(mv_dst - 1*8, s->current_picture.f.motion_val[list][b_xy + 0]); - ref_cache[0 - 1*8]= - ref_cache[1 - 1*8]= ref2frm[list][s->current_picture.f.ref_index[list][b8_xy + 0]]; - ref_cache[2 - 1*8]= - ref_cache[3 - 1*8]= ref2frm[list][s->current_picture.f.ref_index[list][b8_xy + 1]]; - }else{ - AV_ZERO128(mv_dst - 1*8); - AV_WN32A(&ref_cache[0 - 1*8], ((LIST_NOT_USED)&0xFF)*0x01010101u); + if (IS_INTER(mb_type) || IS_DIRECT(mb_type)) { + if (USES_LIST(top_type, list)) { + const int b_xy = h->mb2b_xy[top_xy] + 3 * b_stride; + const int b8_xy = 4 * top_xy + 2; + int (*ref2frm)[64] = h->ref2frm[h->slice_table[top_xy] & (MAX_SLICES - 1)][0] + (MB_MBAFF ? 20 : 2); + AV_COPY128(mv_dst - 1 * 8, s->current_picture.f.motion_val[list][b_xy + 0]); + ref_cache[0 - 1 * 8] = + ref_cache[1 - 1 * 8] = ref2frm[list][s->current_picture.f.ref_index[list][b8_xy + 0]]; + ref_cache[2 - 1 * 8] = + ref_cache[3 - 1 * 8] = ref2frm[list][s->current_picture.f.ref_index[list][b8_xy + 1]]; + } else { + AV_ZERO128(mv_dst - 1 * 8); + AV_WN32A(&ref_cache[0 - 1 * 8], ((LIST_NOT_USED) & 0xFF) * 0x01010101u); } - if(!IS_INTERLACED(mb_type^left_type[LTOP])){ - if(USES_LIST(left_type[LTOP], list)){ - const int b_xy= h->mb2b_xy[left_xy[LTOP]] + 3; - const int b8_xy= 4*left_xy[LTOP] + 1; - int (*ref2frm)[64] = h->ref2frm[ h->slice_table[left_xy[LTOP]]&(MAX_SLICES-1) ][0] + (MB_MBAFF ? 20 : 2); - AV_COPY32(mv_dst - 1 + 0, s->current_picture.f.motion_val[list][b_xy + b_stride*0]); - AV_COPY32(mv_dst - 1 + 8, s->current_picture.f.motion_val[list][b_xy + b_stride*1]); - AV_COPY32(mv_dst - 1 + 16, s->current_picture.f.motion_val[list][b_xy + b_stride*2]); - AV_COPY32(mv_dst - 1 + 24, s->current_picture.f.motion_val[list][b_xy + b_stride*3]); - ref_cache[-1 + 0]= - ref_cache[-1 + 8]= ref2frm[list][s->current_picture.f.ref_index[list][b8_xy + 2*0]]; - ref_cache[-1 + 16]= - ref_cache[-1 + 24]= ref2frm[list][s->current_picture.f.ref_index[list][b8_xy + 2*1]]; - }else{ - AV_ZERO32(mv_dst - 1 + 0); - AV_ZERO32(mv_dst - 1 + 8); - AV_ZERO32(mv_dst - 1 +16); - AV_ZERO32(mv_dst - 1 +24); - ref_cache[-1 + 0]= - ref_cache[-1 + 8]= - ref_cache[-1 + 16]= - ref_cache[-1 + 24]= LIST_NOT_USED; + if (!IS_INTERLACED(mb_type ^ left_type[LTOP])) { + if (USES_LIST(left_type[LTOP], list)) { + const int b_xy = h->mb2b_xy[left_xy[LTOP]] + 3; + const int b8_xy = 4 * left_xy[LTOP] + 1; + int (*ref2frm)[64] = h->ref2frm[h->slice_table[left_xy[LTOP]] & (MAX_SLICES - 1)][0] + (MB_MBAFF ? 20 : 2); + AV_COPY32(mv_dst - 1 + 0, s->current_picture.f.motion_val[list][b_xy + b_stride * 0]); + AV_COPY32(mv_dst - 1 + 8, s->current_picture.f.motion_val[list][b_xy + b_stride * 1]); + AV_COPY32(mv_dst - 1 + 16, s->current_picture.f.motion_val[list][b_xy + b_stride * 2]); + AV_COPY32(mv_dst - 1 + 24, s->current_picture.f.motion_val[list][b_xy + b_stride * 3]); + ref_cache[-1 + 0] = + ref_cache[-1 + 8] = ref2frm[list][s->current_picture.f.ref_index[list][b8_xy + 2 * 0]]; + ref_cache[-1 + 16] = + ref_cache[-1 + 24] = ref2frm[list][s->current_picture.f.ref_index[list][b8_xy + 2 * 1]]; + } else { + AV_ZERO32(mv_dst - 1 + 0); + AV_ZERO32(mv_dst - 1 + 8); + AV_ZERO32(mv_dst - 1 + 16); + AV_ZERO32(mv_dst - 1 + 24); + ref_cache[-1 + 0] = + ref_cache[-1 + 8] = + ref_cache[-1 + 16] = + ref_cache[-1 + 24] = LIST_NOT_USED; } } } - if(!USES_LIST(mb_type, list)){ - fill_rectangle(mv_dst, 4, 4, 8, pack16to32(0,0), 4); - AV_WN32A(&ref_cache[0*8], ((LIST_NOT_USED)&0xFF)*0x01010101u); - AV_WN32A(&ref_cache[1*8], ((LIST_NOT_USED)&0xFF)*0x01010101u); - AV_WN32A(&ref_cache[2*8], ((LIST_NOT_USED)&0xFF)*0x01010101u); - AV_WN32A(&ref_cache[3*8], ((LIST_NOT_USED)&0xFF)*0x01010101u); + if (!USES_LIST(mb_type, list)) { + fill_rectangle(mv_dst, 4, 4, 8, pack16to32(0, 0), 4); + AV_WN32A(&ref_cache[0 * 8], ((LIST_NOT_USED) & 0xFF) * 0x01010101u); + AV_WN32A(&ref_cache[1 * 8], ((LIST_NOT_USED) & 0xFF) * 0x01010101u); + AV_WN32A(&ref_cache[2 * 8], ((LIST_NOT_USED) & 0xFF) * 0x01010101u); + AV_WN32A(&ref_cache[3 * 8], ((LIST_NOT_USED) & 0xFF) * 0x01010101u); return; } { - int8_t *ref = &s->current_picture.f.ref_index[list][4*mb_xy]; - int (*ref2frm)[64] = h->ref2frm[ h->slice_num&(MAX_SLICES-1) ][0] + (MB_MBAFF ? 20 : 2); - uint32_t ref01 = (pack16to32(ref2frm[list][ref[0]],ref2frm[list][ref[1]])&0x00FF00FF)*0x0101; - uint32_t ref23 = (pack16to32(ref2frm[list][ref[2]],ref2frm[list][ref[3]])&0x00FF00FF)*0x0101; - AV_WN32A(&ref_cache[0*8], ref01); - AV_WN32A(&ref_cache[1*8], ref01); - AV_WN32A(&ref_cache[2*8], ref23); - AV_WN32A(&ref_cache[3*8], ref23); + int8_t *ref = &s->current_picture.f.ref_index[list][4 * mb_xy]; + int (*ref2frm)[64] = h->ref2frm[h->slice_num & (MAX_SLICES - 1)][0] + (MB_MBAFF ? 20 : 2); + uint32_t ref01 = (pack16to32(ref2frm[list][ref[0]], ref2frm[list][ref[1]]) & 0x00FF00FF) * 0x0101; + uint32_t ref23 = (pack16to32(ref2frm[list][ref[2]], ref2frm[list][ref[3]]) & 0x00FF00FF) * 0x0101; + AV_WN32A(&ref_cache[0 * 8], ref01); + AV_WN32A(&ref_cache[1 * 8], ref01); + AV_WN32A(&ref_cache[2 * 8], ref23); + AV_WN32A(&ref_cache[3 * 8], ref23); } { - int16_t (*mv_src)[2] = &s->current_picture.f.motion_val[list][4*s->mb_x + 4*s->mb_y*b_stride]; - AV_COPY128(mv_dst + 8*0, mv_src + 0*b_stride); - AV_COPY128(mv_dst + 8*1, mv_src + 1*b_stride); - AV_COPY128(mv_dst + 8*2, mv_src + 2*b_stride); - AV_COPY128(mv_dst + 8*3, mv_src + 3*b_stride); + int16_t(*mv_src)[2] = &s->current_picture.f.motion_val[list][4 * s->mb_x + 4 * s->mb_y * b_stride]; + AV_COPY128(mv_dst + 8 * 0, mv_src + 0 * b_stride); + AV_COPY128(mv_dst + 8 * 1, mv_src + 1 * b_stride); + AV_COPY128(mv_dst + 8 * 2, mv_src + 2 * b_stride); + AV_COPY128(mv_dst + 8 * 3, mv_src + 3 * b_stride); } } @@ -3332,52 +3596,56 @@ static av_always_inline void fill_filter_caches_inter(H264Context *h, MpegEncCon * * @return non zero if the loop filter can be skipped */ -static int fill_filter_caches(H264Context *h, int mb_type){ - MpegEncContext * const s = &h->s; - const int mb_xy= h->mb_xy; +static int fill_filter_caches(H264Context *h, int mb_type) +{ + MpegEncContext *const s = &h->s; + const int mb_xy = h->mb_xy; int top_xy, left_xy[LEFT_MBS]; int top_type, left_type[LEFT_MBS]; uint8_t *nnz; uint8_t *nnz_cache; - top_xy = mb_xy - (s->mb_stride << MB_FIELD); + top_xy = mb_xy - (s->mb_stride << MB_FIELD); /* Wow, what a mess, why didn't they simplify the interlacing & intra * stuff, I can't imagine that these complex rules are worth it. */ - left_xy[LBOT] = left_xy[LTOP] = mb_xy-1; - if(FRAME_MBAFF){ - const int left_mb_field_flag = IS_INTERLACED(s->current_picture.f.mb_type[mb_xy - 1]); - const int curr_mb_field_flag = IS_INTERLACED(mb_type); - if(s->mb_y&1){ - if (left_mb_field_flag != curr_mb_field_flag) { + left_xy[LBOT] = left_xy[LTOP] = mb_xy - 1; + if (FRAME_MBAFF) { + const int left_mb_field_flag = IS_INTERLACED(s->current_picture.f.mb_type[mb_xy - 1]); + const int curr_mb_field_flag = IS_INTERLACED(mb_type); + if (s->mb_y & 1) { + if (left_mb_field_flag != curr_mb_field_flag) left_xy[LTOP] -= s->mb_stride; - } - }else{ - if(curr_mb_field_flag){ - top_xy += s->mb_stride & (((s->current_picture.f.mb_type[top_xy] >> 7) & 1) - 1); - } - if (left_mb_field_flag != curr_mb_field_flag) { + } else { + if (curr_mb_field_flag) + top_xy += s->mb_stride & + (((s->current_picture.f.mb_type[top_xy] >> 7) & 1) - 1); + if (left_mb_field_flag != curr_mb_field_flag) left_xy[LBOT] += s->mb_stride; - } } } - h->top_mb_xy = top_xy; + h->top_mb_xy = top_xy; h->left_mb_xy[LTOP] = left_xy[LTOP]; h->left_mb_xy[LBOT] = left_xy[LBOT]; { - //for sufficiently low qp, filtering wouldn't do anything - //this is a conservative estimate: could also check beta_offset and more accurate chroma_qp - int qp_thresh = h->qp_thresh; //FIXME strictly we should store qp_thresh for each mb of a slice - int qp = s->current_picture.f.qscale_table[mb_xy]; - if(qp <= qp_thresh - && (left_xy[LTOP] < 0 || ((qp + s->current_picture.f.qscale_table[left_xy[LTOP]] + 1) >> 1) <= qp_thresh) - && (top_xy < 0 || ((qp + s->current_picture.f.qscale_table[top_xy ] + 1) >> 1) <= qp_thresh)) { - if(!FRAME_MBAFF) + /* For sufficiently low qp, filtering wouldn't do anything. + * This is a conservative estimate: could also check beta_offset + * and more accurate chroma_qp. */ + int qp_thresh = h->qp_thresh; // FIXME strictly we should store qp_thresh for each mb of a slice + int qp = s->current_picture.f.qscale_table[mb_xy]; + if (qp <= qp_thresh && + (left_xy[LTOP] < 0 || + ((qp + s->current_picture.f.qscale_table[left_xy[LTOP]] + 1) >> 1) <= qp_thresh) && + (top_xy < 0 || + ((qp + s->current_picture.f.qscale_table[top_xy] + 1) >> 1) <= qp_thresh)) { + if (!FRAME_MBAFF) return 1; - if ((left_xy[LTOP] < 0 || ((qp + s->current_picture.f.qscale_table[left_xy[LBOT] ] + 1) >> 1) <= qp_thresh) && - (top_xy < s->mb_stride || ((qp + s->current_picture.f.qscale_table[top_xy - s->mb_stride] + 1) >> 1) <= qp_thresh)) + if ((left_xy[LTOP] < 0 || + ((qp + s->current_picture.f.qscale_table[left_xy[LBOT]] + 1) >> 1) <= qp_thresh) && + (top_xy < s->mb_stride || + ((qp + s->current_picture.f.qscale_table[top_xy - s->mb_stride] + 1) >> 1) <= qp_thresh)) return 1; } } @@ -3385,166 +3653,190 @@ static int fill_filter_caches(H264Context *h, int mb_type){ top_type = s->current_picture.f.mb_type[top_xy]; left_type[LTOP] = s->current_picture.f.mb_type[left_xy[LTOP]]; left_type[LBOT] = s->current_picture.f.mb_type[left_xy[LBOT]]; - if(h->deblocking_filter == 2){ - if(h->slice_table[top_xy ] != h->slice_num) top_type= 0; - if(h->slice_table[left_xy[LBOT]] != h->slice_num) left_type[LTOP]= left_type[LBOT]= 0; - }else{ - if(h->slice_table[top_xy ] == 0xFFFF) top_type= 0; - if(h->slice_table[left_xy[LBOT]] == 0xFFFF) left_type[LTOP]= left_type[LBOT] =0; + if (h->deblocking_filter == 2) { + if (h->slice_table[top_xy] != h->slice_num) + top_type = 0; + if (h->slice_table[left_xy[LBOT]] != h->slice_num) + left_type[LTOP] = left_type[LBOT] = 0; + } else { + if (h->slice_table[top_xy] == 0xFFFF) + top_type = 0; + if (h->slice_table[left_xy[LBOT]] == 0xFFFF) + left_type[LTOP] = left_type[LBOT] = 0; } - h->top_type = top_type; - h->left_type[LTOP]= left_type[LTOP]; - h->left_type[LBOT]= left_type[LBOT]; + h->top_type = top_type; + h->left_type[LTOP] = left_type[LTOP]; + h->left_type[LBOT] = left_type[LBOT]; - if(IS_INTRA(mb_type)) + if (IS_INTRA(mb_type)) return 0; - fill_filter_caches_inter(h, s, mb_type, top_xy, left_xy, top_type, left_type, mb_xy, 0); - if(h->list_count == 2) - fill_filter_caches_inter(h, s, mb_type, top_xy, left_xy, top_type, left_type, mb_xy, 1); + fill_filter_caches_inter(h, s, mb_type, top_xy, left_xy, + top_type, left_type, mb_xy, 0); + if (h->list_count == 2) + fill_filter_caches_inter(h, s, mb_type, top_xy, left_xy, + top_type, left_type, mb_xy, 1); - nnz = h->non_zero_count[mb_xy]; + nnz = h->non_zero_count[mb_xy]; nnz_cache = h->non_zero_count_cache; - AV_COPY32(&nnz_cache[4+8*1], &nnz[ 0]); - AV_COPY32(&nnz_cache[4+8*2], &nnz[ 4]); - AV_COPY32(&nnz_cache[4+8*3], &nnz[ 8]); - AV_COPY32(&nnz_cache[4+8*4], &nnz[12]); - h->cbp= h->cbp_table[mb_xy]; + AV_COPY32(&nnz_cache[4 + 8 * 1], &nnz[0]); + AV_COPY32(&nnz_cache[4 + 8 * 2], &nnz[4]); + AV_COPY32(&nnz_cache[4 + 8 * 3], &nnz[8]); + AV_COPY32(&nnz_cache[4 + 8 * 4], &nnz[12]); + h->cbp = h->cbp_table[mb_xy]; - if(top_type){ + if (top_type) { nnz = h->non_zero_count[top_xy]; - AV_COPY32(&nnz_cache[4+8*0], &nnz[3*4]); + AV_COPY32(&nnz_cache[4 + 8 * 0], &nnz[3 * 4]); } - if(left_type[LTOP]){ + if (left_type[LTOP]) { nnz = h->non_zero_count[left_xy[LTOP]]; - nnz_cache[3+8*1]= nnz[3+0*4]; - nnz_cache[3+8*2]= nnz[3+1*4]; - nnz_cache[3+8*3]= nnz[3+2*4]; - nnz_cache[3+8*4]= nnz[3+3*4]; + nnz_cache[3 + 8 * 1] = nnz[3 + 0 * 4]; + nnz_cache[3 + 8 * 2] = nnz[3 + 1 * 4]; + nnz_cache[3 + 8 * 3] = nnz[3 + 2 * 4]; + nnz_cache[3 + 8 * 4] = nnz[3 + 3 * 4]; } - // CAVLC 8x8dct requires NNZ values for residual decoding that differ from what the loop filter needs - if(!CABAC && h->pps.transform_8x8_mode){ - if(IS_8x8DCT(top_type)){ - nnz_cache[4+8*0]= - nnz_cache[5+8*0]= (h->cbp_table[top_xy] & 0x4000) >> 12; - nnz_cache[6+8*0]= - nnz_cache[7+8*0]= (h->cbp_table[top_xy] & 0x8000) >> 12; + /* CAVLC 8x8dct requires NNZ values for residual decoding that differ + * from what the loop filter needs */ + if (!CABAC && h->pps.transform_8x8_mode) { + if (IS_8x8DCT(top_type)) { + nnz_cache[4 + 8 * 0] = + nnz_cache[5 + 8 * 0] = (h->cbp_table[top_xy] & 0x4000) >> 12; + nnz_cache[6 + 8 * 0] = + nnz_cache[7 + 8 * 0] = (h->cbp_table[top_xy] & 0x8000) >> 12; } - if(IS_8x8DCT(left_type[LTOP])){ - nnz_cache[3+8*1]= - nnz_cache[3+8*2]= (h->cbp_table[left_xy[LTOP]]&0x2000) >> 12; //FIXME check MBAFF + if (IS_8x8DCT(left_type[LTOP])) { + nnz_cache[3 + 8 * 1] = + nnz_cache[3 + 8 * 2] = (h->cbp_table[left_xy[LTOP]] & 0x2000) >> 12; // FIXME check MBAFF } - if(IS_8x8DCT(left_type[LBOT])){ - nnz_cache[3+8*3]= - nnz_cache[3+8*4]= (h->cbp_table[left_xy[LBOT]]&0x8000) >> 12; //FIXME check MBAFF + if (IS_8x8DCT(left_type[LBOT])) { + nnz_cache[3 + 8 * 3] = + nnz_cache[3 + 8 * 4] = (h->cbp_table[left_xy[LBOT]] & 0x8000) >> 12; // FIXME check MBAFF } - if(IS_8x8DCT(mb_type)){ - nnz_cache[scan8[0 ]]= nnz_cache[scan8[1 ]]= - nnz_cache[scan8[2 ]]= nnz_cache[scan8[3 ]]= (h->cbp & 0x1000) >> 12; + if (IS_8x8DCT(mb_type)) { + nnz_cache[scan8[0]] = + nnz_cache[scan8[1]] = + nnz_cache[scan8[2]] = + nnz_cache[scan8[3]] = (h->cbp & 0x1000) >> 12; - nnz_cache[scan8[0+ 4]]= nnz_cache[scan8[1+ 4]]= - nnz_cache[scan8[2+ 4]]= nnz_cache[scan8[3+ 4]]= (h->cbp & 0x2000) >> 12; + nnz_cache[scan8[0 + 4]] = + nnz_cache[scan8[1 + 4]] = + nnz_cache[scan8[2 + 4]] = + nnz_cache[scan8[3 + 4]] = (h->cbp & 0x2000) >> 12; - nnz_cache[scan8[0+ 8]]= nnz_cache[scan8[1+ 8]]= - nnz_cache[scan8[2+ 8]]= nnz_cache[scan8[3+ 8]]= (h->cbp & 0x4000) >> 12; + nnz_cache[scan8[0 + 8]] = + nnz_cache[scan8[1 + 8]] = + nnz_cache[scan8[2 + 8]] = + nnz_cache[scan8[3 + 8]] = (h->cbp & 0x4000) >> 12; - nnz_cache[scan8[0+12]]= nnz_cache[scan8[1+12]]= - nnz_cache[scan8[2+12]]= nnz_cache[scan8[3+12]]= (h->cbp & 0x8000) >> 12; + nnz_cache[scan8[0 + 12]] = + nnz_cache[scan8[1 + 12]] = + nnz_cache[scan8[2 + 12]] = + nnz_cache[scan8[3 + 12]] = (h->cbp & 0x8000) >> 12; } } return 0; } -static void loop_filter(H264Context *h, int start_x, int end_x){ - MpegEncContext * const s = &h->s; - uint8_t *dest_y, *dest_cb, *dest_cr; +static void loop_filter(H264Context *h, int start_x, int end_x) +{ + MpegEncContext *const s = &h->s; + uint8_t *dest_y, *dest_cb, *dest_cr; int linesize, uvlinesize, mb_x, mb_y; - const int end_mb_y= s->mb_y + FRAME_MBAFF; - const int old_slice_type= h->slice_type; - const int pixel_shift = h->pixel_shift; - const int block_h = 16 >> s->chroma_y_shift; + const int end_mb_y = s->mb_y + FRAME_MBAFF; + const int old_slice_type = h->slice_type; + const int pixel_shift = h->pixel_shift; + const int block_h = 16 >> s->chroma_y_shift; - if(h->deblocking_filter) { - for(mb_x= start_x; mb_xdeblocking_filter) { + for (mb_x = start_x; mb_x < end_x; mb_x++) + for (mb_y = end_mb_y - FRAME_MBAFF; mb_y <= end_mb_y; mb_y++) { int mb_xy, mb_type; - mb_xy = h->mb_xy = mb_x + mb_y*s->mb_stride; - h->slice_num= h->slice_table[mb_xy]; - mb_type = s->current_picture.f.mb_type[mb_xy]; - h->list_count= h->list_counts[mb_xy]; + mb_xy = h->mb_xy = mb_x + mb_y * s->mb_stride; + h->slice_num = h->slice_table[mb_xy]; + mb_type = s->current_picture.f.mb_type[mb_xy]; + h->list_count = h->list_counts[mb_xy]; - if(FRAME_MBAFF) - h->mb_mbaff = h->mb_field_decoding_flag = !!IS_INTERLACED(mb_type); + if (FRAME_MBAFF) + h->mb_mbaff = + h->mb_field_decoding_flag = !!IS_INTERLACED(mb_type); - s->mb_x= mb_x; - s->mb_y= mb_y; - dest_y = s->current_picture.f.data[0] + ((mb_x << pixel_shift) + mb_y * s->linesize ) * 16; - dest_cb = s->current_picture.f.data[1] + (mb_x << pixel_shift) * (8 << CHROMA444) + mb_y * s->uvlinesize * block_h; - dest_cr = s->current_picture.f.data[2] + (mb_x << pixel_shift) * (8 << CHROMA444) + mb_y * s->uvlinesize * block_h; - //FIXME simplify above + s->mb_x = mb_x; + s->mb_y = mb_y; + dest_y = s->current_picture.f.data[0] + + ((mb_x << pixel_shift) + mb_y * s->linesize) * 16; + dest_cb = s->current_picture.f.data[1] + + (mb_x << pixel_shift) * (8 << CHROMA444) + + mb_y * s->uvlinesize * block_h; + dest_cr = s->current_picture.f.data[2] + + (mb_x << pixel_shift) * (8 << CHROMA444) + + mb_y * s->uvlinesize * block_h; + // FIXME simplify above if (MB_FIELD) { - linesize = h->mb_linesize = s->linesize * 2; + linesize = h->mb_linesize = s->linesize * 2; uvlinesize = h->mb_uvlinesize = s->uvlinesize * 2; - if(mb_y&1){ //FIXME move out of this function? - dest_y -= s->linesize*15; - dest_cb-= s->uvlinesize * (block_h - 1); - dest_cr-= s->uvlinesize * (block_h - 1); + if (mb_y & 1) { // FIXME move out of this function? + dest_y -= s->linesize * 15; + dest_cb -= s->uvlinesize * (block_h - 1); + dest_cr -= s->uvlinesize * (block_h - 1); } } else { linesize = h->mb_linesize = s->linesize; uvlinesize = h->mb_uvlinesize = s->uvlinesize; } - backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0); - if(fill_filter_caches(h, mb_type)) + backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, + uvlinesize, 0); + if (fill_filter_caches(h, mb_type)) continue; h->chroma_qp[0] = get_chroma_qp(h, 0, s->current_picture.f.qscale_table[mb_xy]); h->chroma_qp[1] = get_chroma_qp(h, 1, s->current_picture.f.qscale_table[mb_xy]); if (FRAME_MBAFF) { - ff_h264_filter_mb (h, mb_x, mb_y, dest_y, dest_cb, dest_cr, linesize, uvlinesize); + ff_h264_filter_mb(h, mb_x, mb_y, dest_y, dest_cb, dest_cr, + linesize, uvlinesize); } else { - ff_h264_filter_mb_fast(h, mb_x, mb_y, dest_y, dest_cb, dest_cr, linesize, uvlinesize); + ff_h264_filter_mb_fast(h, mb_x, mb_y, dest_y, dest_cb, + dest_cr, linesize, uvlinesize); } } - } } - h->slice_type= old_slice_type; - s->mb_x= end_x; - s->mb_y= end_mb_y - FRAME_MBAFF; + h->slice_type = old_slice_type; + s->mb_x = end_x; + s->mb_y = end_mb_y - FRAME_MBAFF; h->chroma_qp[0] = get_chroma_qp(h, 0, s->qscale); h->chroma_qp[1] = get_chroma_qp(h, 1, s->qscale); } -static void predict_field_decoding_flag(H264Context *h){ - MpegEncContext * const s = &h->s; - const int mb_xy= s->mb_x + s->mb_y*s->mb_stride; - int mb_type = (h->slice_table[mb_xy-1] == h->slice_num) - ? s->current_picture.f.mb_type[mb_xy - 1] - : (h->slice_table[mb_xy-s->mb_stride] == h->slice_num) - ? s->current_picture.f.mb_type[mb_xy - s->mb_stride] - : 0; - h->mb_mbaff = h->mb_field_decoding_flag = IS_INTERLACED(mb_type) ? 1 : 0; +static void predict_field_decoding_flag(H264Context *h) +{ + MpegEncContext *const s = &h->s; + const int mb_xy = s->mb_x + s->mb_y * s->mb_stride; + int mb_type = (h->slice_table[mb_xy - 1] == h->slice_num) ? + s->current_picture.f.mb_type[mb_xy - 1] : + (h->slice_table[mb_xy - s->mb_stride] == h->slice_num) ? + s->current_picture.f.mb_type[mb_xy - s->mb_stride] : 0; + h->mb_mbaff = h->mb_field_decoding_flag = IS_INTERLACED(mb_type) ? 1 : 0; } /** * Draw edges and report progress for the last MB row. */ -static void decode_finish_row(H264Context *h){ - MpegEncContext * const s = &h->s; - int top = 16*(s->mb_y >> FIELD_PICTURE); - int height = 16 << FRAME_MBAFF; +static void decode_finish_row(H264Context *h) +{ + MpegEncContext *const s = &h->s; + int top = 16 * (s->mb_y >> FIELD_PICTURE); + int pic_height = 16 * s->mb_height >> FIELD_PICTURE; + int height = 16 << FRAME_MBAFF; int deblock_border = (16 + 4) << FRAME_MBAFF; - int pic_height = 16*s->mb_height >> FIELD_PICTURE; if (h->deblocking_filter) { - if((top + height) >= pic_height) + if ((top + height) >= pic_height) height += deblock_border; - top -= deblock_border; } @@ -3553,144 +3845,175 @@ static void decode_finish_row(H264Context *h){ height = FFMIN(height, pic_height - top); if (top < h->emu_edge_height) { - height = top+height; - top = 0; + height = top + height; + top = 0; } ff_draw_horiz_band(s, top, height); - if (s->dropable) return; + if (s->dropable) + return; ff_thread_report_progress(&s->current_picture_ptr->f, top + height - 1, s->picture_structure == PICT_BOTTOM_FIELD); } -static int decode_slice(struct AVCodecContext *avctx, void *arg){ - H264Context *h = *(void**)arg; - MpegEncContext * const s = &h->s; - const int part_mask= s->partitioned_frame ? (ER_AC_END|ER_AC_ERROR) : 0x7F; +static int decode_slice(struct AVCodecContext *avctx, void *arg) +{ + H264Context *h = *(void **)arg; + MpegEncContext *const s = &h->s; + const int part_mask = s->partitioned_frame ? (ER_AC_END | ER_AC_ERROR) + : 0x7F; int lf_x_start = s->mb_x; - s->mb_skip_run= -1; + s->mb_skip_run = -1; - h->is_complex = FRAME_MBAFF || s->picture_structure != PICT_FRAME || s->codec_id != CODEC_ID_H264 || - (CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY)); + h->is_complex = FRAME_MBAFF || s->picture_structure != PICT_FRAME || + s->codec_id != CODEC_ID_H264 || + (CONFIG_GRAY && (s->flags & CODEC_FLAG_GRAY)); - if( h->pps.cabac ) { + if (h->pps.cabac) { /* realign */ - align_get_bits( &s->gb ); + align_get_bits(&s->gb); /* init cabac */ - ff_init_cabac_states( &h->cabac); - ff_init_cabac_decoder( &h->cabac, - s->gb.buffer + get_bits_count(&s->gb)/8, - (get_bits_left(&s->gb) + 7)/8); + ff_init_cabac_states(&h->cabac); + ff_init_cabac_decoder(&h->cabac, + s->gb.buffer + get_bits_count(&s->gb) / 8, + (get_bits_left(&s->gb) + 7) / 8); ff_h264_init_cabac_states(h); - for(;;){ -//START_TIMER + for (;;) { + // START_TIMER int ret = ff_h264_decode_mb_cabac(h); int eos; -//STOP_TIMER("decode_mb_cabac") + // STOP_TIMER("decode_mb_cabac") - if(ret>=0) ff_h264_hl_decode_mb(h); + if (ret >= 0) + ff_h264_hl_decode_mb(h); - if( ret >= 0 && FRAME_MBAFF ) { //FIXME optimal? or let mb_decode decode 16x32 ? + // FIXME optimal? or let mb_decode decode 16x32 ? + if (ret >= 0 && FRAME_MBAFF) { s->mb_y++; ret = ff_h264_decode_mb_cabac(h); - if(ret>=0) ff_h264_hl_decode_mb(h); + if (ret >= 0) + ff_h264_hl_decode_mb(h); s->mb_y--; } - eos = get_cabac_terminate( &h->cabac ); + eos = get_cabac_terminate(&h->cabac); - if((s->workaround_bugs & FF_BUG_TRUNCATED) && h->cabac.bytestream > h->cabac.bytestream_end + 2){ - ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, ER_MB_END&part_mask); - if (s->mb_x >= lf_x_start) loop_filter(h, lf_x_start, s->mb_x + 1); + if ((s->workaround_bugs & FF_BUG_TRUNCATED) && + h->cabac.bytestream > h->cabac.bytestream_end + 2) { + ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x - 1, + s->mb_y, ER_MB_END & part_mask); + if (s->mb_x >= lf_x_start) + loop_filter(h, lf_x_start, s->mb_x + 1); return 0; } - if( ret < 0 || h->cabac.bytestream > h->cabac.bytestream_end + 2) { - av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding MB %d %d, bytestream (%td)\n", s->mb_x, s->mb_y, h->cabac.bytestream_end - h->cabac.bytestream); - ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR&part_mask); + if (ret < 0 || h->cabac.bytestream > h->cabac.bytestream_end + 2) { + av_log(h->s.avctx, AV_LOG_ERROR, + "error while decoding MB %d %d, bytestream (%td)\n", + s->mb_x, s->mb_y, + h->cabac.bytestream_end - h->cabac.bytestream); + ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, + s->mb_y, ER_MB_ERROR & part_mask); return -1; } - if( ++s->mb_x >= s->mb_width ) { + if (++s->mb_x >= s->mb_width) { loop_filter(h, lf_x_start, s->mb_x); s->mb_x = lf_x_start = 0; decode_finish_row(h); ++s->mb_y; - if(FIELD_OR_MBAFF_PICTURE) { + if (FIELD_OR_MBAFF_PICTURE) { ++s->mb_y; - if(FRAME_MBAFF && s->mb_y < s->mb_height) + if (FRAME_MBAFF && s->mb_y < s->mb_height) predict_field_decoding_flag(h); } } - if( eos || s->mb_y >= s->mb_height ) { - tprintf(s->avctx, "slice end %d %d\n", get_bits_count(&s->gb), s->gb.size_in_bits); - ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, ER_MB_END&part_mask); - if (s->mb_x > lf_x_start) loop_filter(h, lf_x_start, s->mb_x); + if (eos || s->mb_y >= s->mb_height) { + tprintf(s->avctx, "slice end %d %d\n", + get_bits_count(&s->gb), s->gb.size_in_bits); + ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x - 1, + s->mb_y, ER_MB_END & part_mask); + if (s->mb_x > lf_x_start) + loop_filter(h, lf_x_start, s->mb_x); return 0; } } - } else { - for(;;){ + for (;;) { int ret = ff_h264_decode_mb_cavlc(h); - if(ret>=0) ff_h264_hl_decode_mb(h); + if (ret >= 0) + ff_h264_hl_decode_mb(h); - if(ret>=0 && FRAME_MBAFF){ //FIXME optimal? or let mb_decode decode 16x32 ? + // FIXME optimal? or let mb_decode decode 16x32 ? + if (ret >= 0 && FRAME_MBAFF) { s->mb_y++; ret = ff_h264_decode_mb_cavlc(h); - if(ret>=0) ff_h264_hl_decode_mb(h); + if (ret >= 0) + ff_h264_hl_decode_mb(h); s->mb_y--; } - if(ret<0){ - av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding MB %d %d\n", s->mb_x, s->mb_y); - ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR&part_mask); + if (ret < 0) { + av_log(h->s.avctx, AV_LOG_ERROR, + "error while decoding MB %d %d\n", s->mb_x, s->mb_y); + ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, + s->mb_y, ER_MB_ERROR & part_mask); return -1; } - if(++s->mb_x >= s->mb_width){ + if (++s->mb_x >= s->mb_width) { loop_filter(h, lf_x_start, s->mb_x); s->mb_x = lf_x_start = 0; decode_finish_row(h); ++s->mb_y; - if(FIELD_OR_MBAFF_PICTURE) { + if (FIELD_OR_MBAFF_PICTURE) { ++s->mb_y; - if(FRAME_MBAFF && s->mb_y < s->mb_height) + if (FRAME_MBAFF && s->mb_y < s->mb_height) predict_field_decoding_flag(h); } - 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 (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_left(&s->gb) == 0) { - ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, ER_MB_END&part_mask); + ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, + s->mb_x - 1, s->mb_y, + ER_MB_END & part_mask); return 0; - }else{ - ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, ER_MB_END&part_mask); + } else { + ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, + s->mb_x, s->mb_y, + ER_MB_END & part_mask); return -1; } } } - if (get_bits_left(&s->gb) <= 0 && s->mb_skip_run <= 0){ - tprintf(s->avctx, "slice end %d %d\n", get_bits_count(&s->gb), s->gb.size_in_bits); + if (get_bits_left(&s->gb) <= 0 && s->mb_skip_run <= 0) { + tprintf(s->avctx, "slice end %d %d\n", + get_bits_count(&s->gb), s->gb.size_in_bits); if (get_bits_left(&s->gb) == 0) { - ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, ER_MB_END&part_mask); - if (s->mb_x > lf_x_start) loop_filter(h, lf_x_start, s->mb_x); + ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, + s->mb_x - 1, s->mb_y, + ER_MB_END & part_mask); + if (s->mb_x > lf_x_start) + loop_filter(h, lf_x_start, s->mb_x); return 0; - }else{ - ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR&part_mask); + } else { + ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, + s->mb_y, ER_MB_ERROR & part_mask); return -1; } @@ -3705,129 +4028,141 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg){ * @param h h264 master context * @param context_count number of contexts to execute */ -static int execute_decode_slices(H264Context *h, int context_count){ - MpegEncContext * const s = &h->s; - AVCodecContext * const avctx= s->avctx; +static int execute_decode_slices(H264Context *h, int context_count) +{ + MpegEncContext *const s = &h->s; + AVCodecContext *const avctx = s->avctx; H264Context *hx; int i; - if (s->avctx->hwaccel || s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU) + if (s->avctx->hwaccel || + s->avctx->codec->capabilities & CODEC_CAP_HWACCEL_VDPAU) return 0; - if(context_count == 1) { + if (context_count == 1) { return decode_slice(avctx, &h); } else { - for(i = 1; i < context_count; i++) { - hx = h->thread_context[i]; + for (i = 1; i < context_count; i++) { + hx = h->thread_context[i]; hx->s.err_recognition = avctx->err_recognition; - hx->s.error_count = 0; + hx->s.error_count = 0; } - avctx->execute(avctx, decode_slice, - h->thread_context, NULL, context_count, sizeof(void*)); + avctx->execute(avctx, decode_slice, h->thread_context, + NULL, context_count, sizeof(void *)); /* pull back stuff from slices to master context */ - hx = h->thread_context[context_count - 1]; - s->mb_x = hx->s.mb_x; - s->mb_y = hx->s.mb_y; - s->dropable = hx->s.dropable; + hx = h->thread_context[context_count - 1]; + s->mb_x = hx->s.mb_x; + s->mb_y = hx->s.mb_y; + s->dropable = hx->s.dropable; s->picture_structure = hx->s.picture_structure; - for(i = 1; i < context_count; i++) + for (i = 1; i < context_count; i++) h->s.error_count += h->thread_context[i]->s.error_count; } return 0; } - -static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size){ - MpegEncContext * const s = &h->s; - AVCodecContext * const avctx= s->avctx; +static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size) +{ + MpegEncContext *const s = &h->s; + AVCodecContext *const avctx = s->avctx; H264Context *hx; ///< thread context int buf_index; int context_count; int next_avc; int pass = !(avctx->active_thread_type & FF_THREAD_FRAME); - int nals_needed=0; ///< number of NALs that need decoding before the next frame thread starts + int nals_needed = 0; ///< number of NALs that need decoding before the next frame thread starts int nal_index; h->max_contexts = s->slice_context_count; - if(!(s->flags2 & CODEC_FLAG2_CHUNKS)){ + if (!(s->flags2 & CODEC_FLAG2_CHUNKS)) { h->current_slice = 0; if (!s->first_field) - s->current_picture_ptr= NULL; + s->current_picture_ptr = NULL; ff_h264_reset_sei(h); } - for(;pass <= 1;pass++){ - buf_index = 0; + for (; pass <= 1; pass++) { + buf_index = 0; context_count = 0; - next_avc = h->is_avc ? 0 : buf_size; - nal_index = 0; - for(;;){ - int consumed; - int dst_length; - int bit_length; - const uint8_t *ptr; - int i, nalsize = 0; - int err; + next_avc = h->is_avc ? 0 : buf_size; + nal_index = 0; + for (;;) { + int consumed; + int dst_length; + int bit_length; + const uint8_t *ptr; + int i, nalsize = 0; + int err; - if(buf_index >= next_avc) { - if (buf_index >= buf_size - h->nal_length_size) break; - nalsize = 0; - for(i = 0; i < h->nal_length_size; i++) - nalsize = (nalsize << 8) | buf[buf_index++]; - if(nalsize <= 0 || nalsize > buf_size - buf_index){ - av_log(h->s.avctx, AV_LOG_ERROR, "AVC: nal size %d\n", nalsize); - break; - } - next_avc= buf_index + nalsize; - } else { - // start code prefix search - for(; buf_index + 3 < next_avc; buf_index++){ - // This should always succeed in the first iteration. - if(buf[buf_index] == 0 && buf[buf_index+1] == 0 && buf[buf_index+2] == 1) + if (buf_index >= next_avc) { + if (buf_index >= buf_size - h->nal_length_size) break; + nalsize = 0; + for (i = 0; i < h->nal_length_size; i++) + nalsize = (nalsize << 8) | buf[buf_index++]; + if (nalsize <= 0 || nalsize > buf_size - buf_index) { + av_log(h->s.avctx, AV_LOG_ERROR, + "AVC: nal size %d\n", nalsize); + break; + } + next_avc = buf_index + nalsize; + } else { + // start code prefix search + for (; buf_index + 3 < next_avc; buf_index++) + // This should always succeed in the first iteration. + if (buf[buf_index] == 0 && + buf[buf_index + 1] == 0 && + buf[buf_index + 2] == 1) + break; + + if (buf_index + 3 >= buf_size) + break; + + buf_index += 3; + if (buf_index >= next_avc) + continue; } - if(buf_index+3 >= buf_size) break; + hx = h->thread_context[context_count]; - buf_index+=3; - if(buf_index >= next_avc) continue; - } + ptr = ff_h264_decode_nal(hx, buf + buf_index, &dst_length, + &consumed, next_avc - buf_index); + if (ptr == NULL || dst_length < 0) + return -1; + i = buf_index + consumed; + if ((s->workaround_bugs & FF_BUG_AUTODETECT) && i + 3 < next_avc && + buf[i] == 0x00 && buf[i + 1] == 0x00 && + buf[i + 2] == 0x01 && buf[i + 3] == 0xE0) + s->workaround_bugs |= FF_BUG_TRUNCATED; - hx = h->thread_context[context_count]; + if (!(s->workaround_bugs & FF_BUG_TRUNCATED)) + while (ptr[dst_length - 1] == 0 && dst_length > 0) + dst_length--; + bit_length = !dst_length ? 0 + : (8 * dst_length - + ff_h264_decode_rbsp_trailing(h, ptr + dst_length - 1)); - ptr= ff_h264_decode_nal(hx, buf + buf_index, &dst_length, &consumed, next_avc - buf_index); - if (ptr==NULL || dst_length < 0){ - return -1; - } - i= buf_index + consumed; - if((s->workaround_bugs & FF_BUG_AUTODETECT) && i+3workaround_bugs |= FF_BUG_TRUNCATED; + 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); - if(!(s->workaround_bugs & FF_BUG_TRUNCATED)){ - while(ptr[dst_length - 1] == 0 && dst_length > 0) - dst_length--; - } - bit_length= !dst_length ? 0 : (8*dst_length - ff_h264_decode_rbsp_trailing(h, ptr + dst_length - 1)); + if (h->is_avc && (nalsize != consumed) && nalsize) + av_log(h->s.avctx, AV_LOG_DEBUG, + "AVC: Consumed only %d bytes instead of %d\n", + consumed, nalsize); - 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); - } + buf_index += consumed; + nal_index++; - if (h->is_avc && (nalsize != consumed) && nalsize){ - av_log(h->s.avctx, AV_LOG_DEBUG, "AVC: Consumed only %d bytes instead of %d\n", consumed, nalsize); - } - - buf_index += consumed; - nal_index++; - - if(pass == 0) { - // packets can sometimes contain multiple PPS/SPS - // e.g. two PAFF field pictures in one packet, or a demuxer which splits NALs strangely - // if so, when frame threading we can't start the next thread until we've read all of them - switch (hx->nal_unit_type) { + if (pass == 0) { + /* packets can sometimes contain multiple PPS/SPS, + * e.g. two PAFF field pictures in one packet, or a demuxer + * which splits NALs strangely if so, when frame threading we + * can't start the next thread until we've read all of them */ + switch (hx->nal_unit_type) { case NAL_SPS: case NAL_PPS: nals_needed = nal_index; @@ -3837,167 +4172,185 @@ static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size){ init_get_bits(&hx->s.gb, ptr, bit_length); if (!get_ue_golomb(&hx->s.gb)) nals_needed = nal_index; + } + continue; } - continue; - } - //FIXME do not discard SEI id - if(avctx->skip_frame >= AVDISCARD_NONREF && h->nal_ref_idc == 0) - continue; + // FIXME do not discard SEI id + if (avctx->skip_frame >= AVDISCARD_NONREF && h->nal_ref_idc == 0) + continue; - again: - err = 0; - switch(hx->nal_unit_type){ - case NAL_IDR_SLICE: - if (h->nal_unit_type != NAL_IDR_SLICE) { - av_log(h->s.avctx, AV_LOG_ERROR, "Invalid mix of idr and non-idr slices"); - return -1; - } - idr(h); // FIXME ensure we don't lose some frames if there is reordering - case NAL_SLICE: - init_get_bits(&hx->s.gb, ptr, bit_length); - hx->intra_gb_ptr= - hx->inter_gb_ptr= &hx->s.gb; - hx->s.data_partitioning = 0; +again: + err = 0; + switch (hx->nal_unit_type) { + case NAL_IDR_SLICE: + if (h->nal_unit_type != NAL_IDR_SLICE) { + av_log(h->s.avctx, AV_LOG_ERROR, + "Invalid mix of idr and non-idr slices"); + return -1; + } + idr(h); // FIXME ensure we don't lose some frames if there is reordering + case NAL_SLICE: + init_get_bits(&hx->s.gb, ptr, bit_length); + hx->intra_gb_ptr = + hx->inter_gb_ptr = &hx->s.gb; + hx->s.data_partitioning = 0; - if((err = decode_slice_header(hx, h))) - break; + if ((err = decode_slice_header(hx, h))) + break; - s->current_picture_ptr->f.key_frame |= + s->current_picture_ptr->f.key_frame |= (hx->nal_unit_type == NAL_IDR_SLICE) || (h->sei_recovery_frame_cnt >= 0); - if (h->current_slice == 1) { - if(!(s->flags2 & CODEC_FLAG2_CHUNKS)) { - decode_postinit(h, nal_index >= nals_needed); - } + if (h->current_slice == 1) { + if (!(s->flags2 & CODEC_FLAG2_CHUNKS)) + decode_postinit(h, nal_index >= nals_needed); - if (s->avctx->hwaccel && s->avctx->hwaccel->start_frame(s->avctx, NULL, 0) < 0) - return -1; - if(CONFIG_H264_VDPAU_DECODER && s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU) - ff_vdpau_h264_picture_start(s); - } - - if(hx->redundant_pic_count==0 - && (avctx->skip_frame < AVDISCARD_NONREF || hx->nal_ref_idc) - && (avctx->skip_frame < AVDISCARD_BIDIR || hx->slice_type_nos!=AV_PICTURE_TYPE_B) - && (avctx->skip_frame < AVDISCARD_NONKEY || hx->slice_type_nos==AV_PICTURE_TYPE_I) - && avctx->skip_frame < AVDISCARD_ALL){ - if(avctx->hwaccel) { - if (avctx->hwaccel->decode_slice(avctx, &buf[buf_index - consumed], consumed) < 0) + if (s->avctx->hwaccel && + s->avctx->hwaccel->start_frame(s->avctx, NULL, 0) < 0) return -1; - }else - if(CONFIG_H264_VDPAU_DECODER && s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU){ - static const uint8_t start_code[] = {0x00, 0x00, 0x01}; - ff_vdpau_add_data_chunk(s, start_code, sizeof(start_code)); - ff_vdpau_add_data_chunk(s, &buf[buf_index - consumed], consumed ); - }else - context_count++; - } - break; - case NAL_DPA: - init_get_bits(&hx->s.gb, ptr, bit_length); - hx->intra_gb_ptr= - hx->inter_gb_ptr= NULL; - - if ((err = decode_slice_header(hx, h)) < 0) - break; - - hx->s.data_partitioning = 1; - - break; - case NAL_DPB: - init_get_bits(&hx->intra_gb, ptr, bit_length); - hx->intra_gb_ptr= &hx->intra_gb; - break; - case NAL_DPC: - init_get_bits(&hx->inter_gb, ptr, bit_length); - hx->inter_gb_ptr= &hx->inter_gb; - - if(hx->redundant_pic_count==0 && hx->intra_gb_ptr && hx->s.data_partitioning - && s->context_initialized - && (avctx->skip_frame < AVDISCARD_NONREF || hx->nal_ref_idc) - && (avctx->skip_frame < AVDISCARD_BIDIR || hx->slice_type_nos!=AV_PICTURE_TYPE_B) - && (avctx->skip_frame < AVDISCARD_NONKEY || hx->slice_type_nos==AV_PICTURE_TYPE_I) - && avctx->skip_frame < AVDISCARD_ALL) - context_count++; - break; - case NAL_SEI: - init_get_bits(&s->gb, ptr, bit_length); - ff_h264_decode_sei(h); - break; - case NAL_SPS: - init_get_bits(&s->gb, ptr, bit_length); - if (ff_h264_decode_seq_parameter_set(h) < 0 && - h->is_avc && (nalsize != consumed) && nalsize) { - av_log(h->s.avctx, AV_LOG_DEBUG, "SPS decoding failure, " - "try parsing the coomplete NAL\n"); - init_get_bits(&s->gb, buf + buf_index + 1 - consumed, - 8 * (nalsize - 1)); - ff_h264_decode_seq_parameter_set(h); - } - - if (s->flags& CODEC_FLAG_LOW_DELAY || - (h->sps.bitstream_restriction_flag && !h->sps.num_reorder_frames)) - s->low_delay=1; - - 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; - ff_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; + if (CONFIG_H264_VDPAU_DECODER && + s->avctx->codec->capabilities & CODEC_CAP_HWACCEL_VDPAU) + ff_vdpau_h264_picture_start(s); } + + if (hx->redundant_pic_count == 0 && + (avctx->skip_frame < AVDISCARD_NONREF || + hx->nal_ref_idc) && + (avctx->skip_frame < AVDISCARD_BIDIR || + hx->slice_type_nos != AV_PICTURE_TYPE_B) && + (avctx->skip_frame < AVDISCARD_NONKEY || + hx->slice_type_nos == AV_PICTURE_TYPE_I) && + avctx->skip_frame < AVDISCARD_ALL) { + if (avctx->hwaccel) { + if (avctx->hwaccel->decode_slice(avctx, + &buf[buf_index - consumed], + consumed) < 0) + return -1; + } else if (CONFIG_H264_VDPAU_DECODER && + s->avctx->codec->capabilities & CODEC_CAP_HWACCEL_VDPAU) { + static const uint8_t start_code[] = { + 0x00, 0x00, 0x01 }; + ff_vdpau_add_data_chunk(s, start_code, + sizeof(start_code)); + ff_vdpau_add_data_chunk(s, &buf[buf_index - consumed], + consumed); + } else + context_count++; + } + break; + case NAL_DPA: + init_get_bits(&hx->s.gb, ptr, bit_length); + hx->intra_gb_ptr = + hx->inter_gb_ptr = NULL; + + if ((err = decode_slice_header(hx, h)) < 0) + break; + + hx->s.data_partitioning = 1; + break; + case NAL_DPB: + init_get_bits(&hx->intra_gb, ptr, bit_length); + hx->intra_gb_ptr = &hx->intra_gb; + break; + case NAL_DPC: + init_get_bits(&hx->inter_gb, ptr, bit_length); + hx->inter_gb_ptr = &hx->inter_gb; + + if (hx->redundant_pic_count == 0 && + hx->intra_gb_ptr && + hx->s.data_partitioning && + s->context_initialized && + (avctx->skip_frame < AVDISCARD_NONREF || hx->nal_ref_idc) && + (avctx->skip_frame < AVDISCARD_BIDIR || + hx->slice_type_nos != AV_PICTURE_TYPE_B) && + (avctx->skip_frame < AVDISCARD_NONKEY || + hx->slice_type_nos == AV_PICTURE_TYPE_I) && + avctx->skip_frame < AVDISCARD_ALL) + context_count++; + break; + case NAL_SEI: + init_get_bits(&s->gb, ptr, bit_length); + ff_h264_decode_sei(h); + break; + case NAL_SPS: + init_get_bits(&s->gb, ptr, bit_length); + if (ff_h264_decode_seq_parameter_set(h) < 0 && + h->is_avc && (nalsize != consumed) && nalsize) { + av_log(h->s.avctx, AV_LOG_DEBUG, + "SPS decoding failure, try parsing the coomplete NAL\n"); + init_get_bits(&s->gb, buf + buf_index + 1 - consumed, + 8 * (nalsize - 1)); + ff_h264_decode_seq_parameter_set(h); + } + + if (s->flags & CODEC_FLAG_LOW_DELAY || + (h->sps.bitstream_restriction_flag && + !h->sps.num_reorder_frames)) + s->low_delay = 1; + + 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; + ff_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); + ff_h264_decode_picture_parameter_set(h, bit_length); + break; + case NAL_AUD: + case NAL_END_SEQUENCE: + case NAL_END_STREAM: + case NAL_FILLER_DATA: + case NAL_SPS_EXT: + case NAL_AUXILIARY_SLICE: + break; + default: + av_log(avctx, AV_LOG_DEBUG, "Unknown NAL code: %d (%d bits)\n", + hx->nal_unit_type, bit_length); } - break; - case NAL_PPS: - init_get_bits(&s->gb, ptr, bit_length); - ff_h264_decode_picture_parameter_set(h, bit_length); + if (context_count == h->max_contexts) { + execute_decode_slices(h, context_count); + context_count = 0; + } - break; - case NAL_AUD: - case NAL_END_SEQUENCE: - case NAL_END_STREAM: - case NAL_FILLER_DATA: - case NAL_SPS_EXT: - case NAL_AUXILIARY_SLICE: - break; - default: - av_log(avctx, AV_LOG_DEBUG, "Unknown NAL code: %d (%d bits)\n", hx->nal_unit_type, bit_length); - } - - if(context_count == h->max_contexts) { - execute_decode_slices(h, context_count); - context_count = 0; - } - - if (err < 0) - av_log(h->s.avctx, AV_LOG_ERROR, "decode_slice_header error\n"); - else if(err == 1) { - /* Slice could not be decoded in parallel mode, copy down - * NAL unit stuff to context 0 and restart. Note that - * rbsp_buffer is not transferred, but since we no longer - * run in parallel mode this should not be an issue. */ - h->nal_unit_type = hx->nal_unit_type; - h->nal_ref_idc = hx->nal_ref_idc; - hx = h; - goto again; + if (err < 0) + av_log(h->s.avctx, AV_LOG_ERROR, "decode_slice_header error\n"); + else if (err == 1) { + /* Slice could not be decoded in parallel mode, copy down + * NAL unit stuff to context 0 and restart. Note that + * rbsp_buffer is not transferred, but since we no longer + * run in parallel mode this should not be an issue. */ + h->nal_unit_type = hx->nal_unit_type; + h->nal_ref_idc = hx->nal_ref_idc; + hx = h; + goto again; + } } } - } - if(context_count) + if (context_count) execute_decode_slices(h, context_count); return buf_index; } @@ -4005,48 +4358,54 @@ static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size){ /** * Return the number of bytes consumed for building the current frame. */ -static int get_consumed_bytes(MpegEncContext *s, int pos, int buf_size){ - if(pos==0) pos=1; //avoid infinite loops (i doubt that is needed but ...) - if(pos+10>buf_size) pos=buf_size; // oops ;) +static int get_consumed_bytes(MpegEncContext *s, int pos, int buf_size) +{ + if (pos == 0) + pos = 1; // avoid infinite loops (i doubt that is needed but ...) + if (pos + 10 > buf_size) + pos = buf_size; // oops ;) - return pos; + return pos; } -static int decode_frame(AVCodecContext *avctx, - void *data, int *data_size, - AVPacket *avpkt) +static int decode_frame(AVCodecContext *avctx, void *data, + int *data_size, AVPacket *avpkt) { const uint8_t *buf = avpkt->data; - int buf_size = avpkt->size; - H264Context *h = avctx->priv_data; - MpegEncContext *s = &h->s; - AVFrame *pict = data; - int buf_index = 0; + int buf_size = avpkt->size; + H264Context *h = avctx->priv_data; + MpegEncContext *s = &h->s; + AVFrame *pict = data; + int buf_index = 0; - s->flags= avctx->flags; - s->flags2= avctx->flags2; + s->flags = avctx->flags; + s->flags2 = avctx->flags2; - /* end of stream, output what is still in the buffers */ - out: + /* end of stream, output what is still in the buffers */ +out: if (buf_size == 0) { Picture *out; int i, out_idx; s->current_picture_ptr = NULL; -//FIXME factorize this with the output code below - out = h->delayed_pic[0]; + // FIXME factorize this with the output code below + out = h->delayed_pic[0]; out_idx = 0; - for (i = 1; h->delayed_pic[i] && !h->delayed_pic[i]->f.key_frame && !h->delayed_pic[i]->mmco_reset; i++) - if(h->delayed_pic[i]->poc < out->poc){ - out = h->delayed_pic[i]; + for (i = 1; + h->delayed_pic[i] && + !h->delayed_pic[i]->f.key_frame && + !h->delayed_pic[i]->mmco_reset; + i++) + if (h->delayed_pic[i]->poc < out->poc) { + out = h->delayed_pic[i]; out_idx = i; } - for(i=out_idx; h->delayed_pic[i]; i++) - h->delayed_pic[i] = h->delayed_pic[i+1]; + for (i = out_idx; h->delayed_pic[i]; i++) + h->delayed_pic[i] = h->delayed_pic[i + 1]; - if(out){ + if (out) { *data_size = sizeof(AVFrame); *pict = out->f; } @@ -4054,8 +4413,8 @@ static int decode_frame(AVCodecContext *avctx, return buf_index; } - buf_index=decode_nal_units(h, buf, buf_size); - if(buf_index < 0) + buf_index = decode_nal_units(h, buf, buf_size); + if (buf_index < 0) return -1; if (!s->current_picture_ptr && h->nal_unit_type == NAL_END_SEQUENCE) { @@ -4063,23 +4422,23 @@ static int decode_frame(AVCodecContext *avctx, goto out; } - if(!(s->flags2 & CODEC_FLAG2_CHUNKS) && !s->current_picture_ptr){ + if (!(s->flags2 & CODEC_FLAG2_CHUNKS) && !s->current_picture_ptr) { if (avctx->skip_frame >= AVDISCARD_NONREF) return 0; av_log(avctx, AV_LOG_ERROR, "no frame!\n"); return -1; } - if(!(s->flags2 & CODEC_FLAG2_CHUNKS) || (s->mb_y >= s->mb_height && s->mb_height)){ - - if(s->flags2 & CODEC_FLAG2_CHUNKS) decode_postinit(h, 1); + if (!(s->flags2 & CODEC_FLAG2_CHUNKS) || + (s->mb_y >= s->mb_height && s->mb_height)) { + if (s->flags2 & CODEC_FLAG2_CHUNKS) + decode_postinit(h, 1); field_end(h, 0); if (!h->next_output_pic) { /* Wait for second field. */ *data_size = 0; - } else { *data_size = sizeof(AVFrame); *pict = h->next_output_pic->f; @@ -4088,53 +4447,34 @@ static int decode_frame(AVCodecContext *avctx, assert(pict->data[0] || !*data_size); ff_print_debug_info(s, pict); -//printf("out %d\n", (int)pict->data[0]); + // printf("out %d\n", (int)pict->data[0]); return get_consumed_bytes(s, buf_index, buf_size); } -#if 0 -static inline void fill_mb_avail(H264Context *h){ - MpegEncContext * const s = &h->s; - const int mb_xy= s->mb_x + s->mb_y*s->mb_stride; - - if(s->mb_y){ - h->mb_avail[0]= s->mb_x && h->slice_table[mb_xy - s->mb_stride - 1] == h->slice_num; - h->mb_avail[1]= h->slice_table[mb_xy - s->mb_stride ] == h->slice_num; - h->mb_avail[2]= s->mb_x+1 < s->mb_width && h->slice_table[mb_xy - s->mb_stride + 1] == h->slice_num; - }else{ - h->mb_avail[0]= - h->mb_avail[1]= - h->mb_avail[2]= 0; - } - h->mb_avail[3]= s->mb_x && h->slice_table[mb_xy - 1] == h->slice_num; - h->mb_avail[4]= 1; //FIXME move out - h->mb_avail[5]= 0; //FIXME move out -} -#endif av_cold void ff_h264_free_context(H264Context *h) { int i; - free_tables(h, 1); //FIXME cleanup init stuff perhaps + free_tables(h, 1); // FIXME cleanup init stuff perhaps - for(i = 0; i < MAX_SPS_COUNT; i++) + for (i = 0; i < MAX_SPS_COUNT; i++) av_freep(h->sps_buffers + i); - for(i = 0; i < MAX_PPS_COUNT; i++) + for (i = 0; i < MAX_PPS_COUNT; i++) av_freep(h->pps_buffers + i); } av_cold int ff_h264_decode_end(AVCodecContext *avctx) { - H264Context *h = avctx->priv_data; + H264Context *h = avctx->priv_data; MpegEncContext *s = &h->s; ff_h264_free_context(h); ff_MPV_common_end(s); -// memset(h, 0, sizeof(H264Context)); + // memset(h, 0, sizeof(H264Context)); return 0; } @@ -4157,20 +4497,21 @@ static const AVProfile profiles[] = { }; AVCodec ff_h264_decoder = { - .name = "h264", - .type = AVMEDIA_TYPE_VIDEO, - .id = CODEC_ID_H264, - .priv_data_size = sizeof(H264Context), - .init = ff_h264_decode_init, - .close = ff_h264_decode_end, - .decode = decode_frame, - .capabilities = /*CODEC_CAP_DRAW_HORIZ_BAND |*/ CODEC_CAP_DR1 | CODEC_CAP_DELAY | - CODEC_CAP_SLICE_THREADS | CODEC_CAP_FRAME_THREADS, - .flush= flush_dpb, - .long_name = NULL_IF_CONFIG_SMALL("H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"), + .name = "h264", + .type = AVMEDIA_TYPE_VIDEO, + .id = CODEC_ID_H264, + .priv_data_size = sizeof(H264Context), + .init = ff_h264_decode_init, + .close = ff_h264_decode_end, + .decode = decode_frame, + .capabilities = /*CODEC_CAP_DRAW_HORIZ_BAND |*/ CODEC_CAP_DR1 | + CODEC_CAP_DELAY | CODEC_CAP_SLICE_THREADS | + CODEC_CAP_FRAME_THREADS, + .flush = flush_dpb, + .long_name = NULL_IF_CONFIG_SMALL("H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"), .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), + .profiles = NULL_IF_CONFIG_SMALL(profiles), }; #if CONFIG_H264_VDPAU_DECODER @@ -4183,9 +4524,10 @@ AVCodec ff_h264_vdpau_decoder = { .close = ff_h264_decode_end, .decode = decode_frame, .capabilities = CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU, - .flush= flush_dpb, - .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), + .flush = flush_dpb, + .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), }; #endif