December 2016 - ffmpeg-cvslog

avcodec/magicyuv: export colorspace and color_range for YUV
by git＠videolan.org 26 Dec '16

26 Dec '16

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Mon Dec 26 17:34:57 2016 +0100| [12461636ea8c2fd014bc835e1d95129f03ecf182] | committer: Paul B Mahol avcodec/magicyuv: export colorspace and color_range for YUV Signed-off-by: Paul B Mahol <onemda(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=12461636ea8c2fd01… --- libavcodec/magicyuv.c | 18 ++++++++++++++++-- 1 file changed, 16 insertions(+), 2 deletions(-) diff --git a/libavcodec/magicyuv.c b/libavcodec/magicyuv.c index 16d7027..526df6f 100644 --- a/libavcodec/magicyuv.c +++ b/libavcodec/magicyuv.c @@ -57,6 +57,8 @@ typedef struct MagicYUVContext { int nb_slices; int planes; // number of encoded planes in bitstream int decorrelate; // postprocessing work + int color_matrix; // video color matrix + int flags; int interlaced; // video is interlaced uint8_t *buf; // pointer to AVPacket->data int hshift[4]; @@ -568,8 +570,10 @@ static int magy_decode_frame(AVCodecContext *avctx, void *data, } s->planes = av_pix_fmt_count_planes(avctx->pix_fmt); - bytestream2_skip(&gbyte, 2); - s->interlaced = !!(bytestream2_get_byte(&gbyte) & 2); + bytestream2_skip(&gbyte, 1); + s->color_matrix = bytestream2_get_byte(&gbyte); + s->flags = bytestream2_get_byte(&gbyte); + s->interlaced = !!(s->flags & 2); bytestream2_skip(&gbyte, 3); width = bytestream2_get_le32(&gbyte); @@ -659,6 +663,16 @@ static int magy_decode_frame(AVCodecContext *avctx, void *data, avctx->pix_fmt == AV_PIX_FMT_GBRAP10) { FFSWAP(uint8_t*, p->data[0], p->data[1]); FFSWAP(int, p->linesize[0], p->linesize[1]); + } else { + switch (s->color_matrix) { + case 1: + p->colorspace = AVCOL_SPC_BT470BG; + break; + case 2: + p->colorspace = AVCOL_SPC_BT709; + break; + } + p->color_range = (s->flags & 4) ? AVCOL_RANGE_JPEG : AVCOL_RANGE_MPEG; } *got_frame = 1;

1 0

NVENC: Update check for Lookahead
by git＠videolan.org 26 Dec '16

26 Dec '16

ffmpeg | branch: master | Ruta Gadkari <rgadkari(a)nvidia.com> | Mon Dec 26 13:37:16 2016 +0530| [67db4ff3b66e96a20ddf1c264d07e146334ae88e] | committer: James Almer NVENC: Update check for Lookahead Reviewed-by: Timo Rothenpieler <timo(a)rothenpieler.org> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=67db4ff3b66e96a20… --- libavcodec/nvenc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libavcodec/nvenc.c b/libavcodec/nvenc.c index 476a53e..f64fd8a 100644 --- a/libavcodec/nvenc.c +++ b/libavcodec/nvenc.c @@ -695,7 +695,7 @@ static av_cold void nvenc_setup_rate_control(AVCodecContext *avctx) av_log(avctx, AV_LOG_VERBOSE, "Temporal AQ enabled.\n"); } - if (ctx->rc_lookahead) { + if (ctx->rc_lookahead > 0) { int lkd_bound = FFMIN(ctx->nb_surfaces, ctx->async_depth) - ctx->encode_config.frameIntervalP - 4;

1 0

avcodec/lossless_videodsp: fix output of add_hfyu_left_pred_int16_c()
by git＠videolan.org 26 Dec '16

26 Dec '16

ffmpeg | branch: master | James Almer <jamrial(a)gmail.com> | Sun Dec 25 23:16:58 2016 -0300| [c3d822855cb5220e478d8b0ba0bafde18aa2f122] | committer: James Almer avcodec/lossless_videodsp: fix output of add_hfyu_left_pred_int16_c() It is now bitexact with the ssse3 and sse4.1 versions of the function. Reviewed-by: Paul B Mahol <onemda(a)gmail.com> Signed-off-by: James Almer <jamrial(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=c3d822855cb5220e4… --- libavcodec/lossless_videodsp.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/libavcodec/lossless_videodsp.c b/libavcodec/lossless_videodsp.c index 3491621..231c25f 100644 --- a/libavcodec/lossless_videodsp.c +++ b/libavcodec/lossless_videodsp.c @@ -100,15 +100,15 @@ static int add_hfyu_left_pred_int16_c(uint16_t *dst, const uint16_t *src, unsign for(i=0; i<w-1; i++){ acc+= src[i]; - dst[i]= acc & mask; + dst[i]= acc &= mask; i++; acc+= src[i]; - dst[i]= acc & mask; + dst[i]= acc &= mask; } for(; i<w; i++){ acc+= src[i]; - dst[i]= acc & mask; + dst[i]= acc &= mask; } return acc;

1 0

avformat/hlsenc: detecting duplicated segment filenames
by git＠videolan.org 26 Dec '16

26 Dec '16

ffmpeg | branch: master | Bela Bodecs <bodecsb(a)vivanet.hu> | Mon Dec 26 18:14:40 2016 +0800| [e7fbd7018932aa349bf52a2bc8e1acd6da058a1b] | committer: Bela Bodecs avformat/hlsenc: detecting duplicated segment filenames ffmpeg-devel with use_localtime parameter hlsenc may produce identical filenames for different but still existing segments. It happens when hls_segment_filename contains syntacticaly correct but inadequate format parameters. Currently there is no any log message when such a situaton occurs but these cases should be avoided in most times. This patch generate warning log messages in these cases. ticketID: #6043 Signed-off-by: Bela Bodecs <bodecsb(a)vivanet.hu> Signed-off-by: Steven Liu <lingjiujianke(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=e7fbd7018932aa349… --- libavformat/hlsenc.c | 36 ++++++++++++++++++++++++++++++++++++ 1 file changed, 36 insertions(+) diff --git a/libavformat/hlsenc.c b/libavformat/hlsenc.c index acf3a30..e46e9b4 100644 --- a/libavformat/hlsenc.c +++ b/libavformat/hlsenc.c @@ -653,6 +653,38 @@ fail: return ret; } +static HLSSegment *find_segment_by_filename(HLSSegment *segment, const char *filename) +{ + /* filename may contain rel/abs path, but segments store only basename */ + char *p = NULL, *dirname = NULL, *path = NULL; + int path_size; + HLSSegment *ret_segment = NULL; + dirname = av_strdup(filename); + if (!dirname) + return NULL; + p = (char *)av_basename(dirname); // av_dirname would return . in case of no dir + *p = '\0'; // maybe empty + + while (segment) { + path_size = strlen(dirname) + strlen(segment->filename) + 1; + path = av_malloc(path_size); + if (!path) + goto end; + av_strlcpy(path, dirname, path_size); + av_strlcat(path, segment->filename, path_size); + if (!strcmp(path,filename)) { + ret_segment = segment; + av_free(path); + goto end; + } + av_free(path); + segment = segment->next; + } +end: + av_free(dirname); + return ret_segment; +} + static int hls_start(AVFormatContext *s) { HLSContext *c = s->priv_data; @@ -686,6 +718,10 @@ static int hls_start(AVFormatContext *s) return AVERROR(EINVAL); } + if (find_segment_by_filename(c->segments, oc->filename) + || find_segment_by_filename(c->old_segments, oc->filename)) { + av_log(c, AV_LOG_WARNING, "Duplicated segment filename detected: %s\n", oc->filename); + } if (c->use_localtime_mkdir) { const char *dir; char *fn_copy = av_strdup(oc->filename);

1 0

avcodec/error_resilience: Move variable initialization down, remove unneeded inits
by git＠videolan.org 25 Dec '16

25 Dec '16

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Sun Dec 25 21:21:31 2016 +0100| [6e26b6e43fec8032662e39c9d056b2d991f89ab4] | committer: Michael Niedermayer avcodec/error_resilience: Move variable initialization down, remove unneeded inits This makes the code faster and easier to read Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=6e26b6e43fec80326… --- libavcodec/error_resilience.c | 21 ++++++++++++++------- 1 file changed, 14 insertions(+), 7 deletions(-) diff --git a/libavcodec/error_resilience.c b/libavcodec/error_resilience.c index 741a9c8..c7dbe17 100644 --- a/libavcodec/error_resilience.c +++ b/libavcodec/error_resilience.c @@ -452,14 +452,14 @@ static void guess_mv(ERContext *s) for (mb_y = 0; mb_y < mb_height; mb_y++) { for (mb_x = (mb_y ^ pass) & 1; mb_x < s->mb_width; mb_x+=2) { const int mb_xy = mb_x + mb_y * s->mb_stride; - int mv_predictor[8][2] = { { 0 } }; - int ref[8] = { 0 }; - int pred_count = 0; + int mv_predictor[8][2]; + int ref[8]; + int pred_count; int j; - int best_score = 256 * 256 * 256 * 64; - int best_pred = 0; - const int mot_index = (mb_x + mb_y * mot_stride) * mot_step; - int prev_x = 0, prev_y = 0, prev_ref = 0; + int best_score; + int best_pred; + int mot_index; + int prev_x, prev_y, prev_ref; if (fixed[mb_xy] == MV_FROZEN) continue; @@ -483,6 +483,8 @@ static void guess_mv(ERContext *s) continue; none_left = 0; + pred_count = 0; + mot_index = (mb_x + mb_y * mot_stride) * mot_step; if (mb_x > 0 && fixed[mb_xy - 1]) { mv_predictor[pred_count][0] = @@ -569,6 +571,9 @@ static void guess_mv(ERContext *s) skip_mean_and_median: /* zero MV */ + mv_predictor[pred_count][0] = + mv_predictor[pred_count][1] = + ref[pred_count] = 0; pred_count++; prev_x = s->cur_pic.motion_val[0][mot_index][0]; @@ -581,6 +586,8 @@ skip_mean_and_median: ref[pred_count] = prev_ref; pred_count++; + best_pred = 0; + best_score = 256 * 256 * 256 * 64; for (j = 0; j < pred_count; j++) { int *linesize = s->cur_pic.f->linesize; int score = 0;

1 0

ffplay: add startup volume option
by git＠videolan.org 25 Dec '16

25 Dec '16

ffmpeg | branch: master | Ganesh Ajjanagadde <gajjanag(a)alum.mit.edu> | Thu Dec 22 09:51:31 2016 -0800| [7b557bf63ff8549f68cd6a53adb78bf1954187c7] | committer: Ganesh Ajjanagadde ffplay: add startup volume option Fixes Ticket 5389. Reviewed-by: Marton Balint <cus(a)passwd.hu> Signed-off-by: Ganesh Ajjanagadde <gajjanag(a)alum.mit.edu> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=7b557bf63ff8549f6… --- doc/ffplay.texi | 4 ++++ ffplay.c | 10 +++++++++- 2 files changed, 13 insertions(+), 1 deletion(-) diff --git a/doc/ffplay.texi b/doc/ffplay.texi index 073b457..378a229 100644 --- a/doc/ffplay.texi +++ b/doc/ffplay.texi @@ -62,6 +62,10 @@ see @ref{time duration syntax,,the Time duration section in the ffmpeg-utils(1) Seek by bytes. @item -nodisp Disable graphical display. +@item -volume +Set the startup volume. 0 means silence, 100 means no volume reduction or +amplification. Negative values are treated as 0, values above 100 are treated +as 100. @item -f @var{fmt} Force format. @item -window_title @var{title} diff --git a/ffplay.c b/ffplay.c index 911fd7f..bca2d5c 100644 --- a/ffplay.c +++ b/ffplay.c @@ -322,6 +322,7 @@ static int subtitle_disable; static const char* wanted_stream_spec[AVMEDIA_TYPE_NB] = {0}; static int seek_by_bytes = -1; static int display_disable; +static int startup_volume = 100; static int show_status = 1; static int av_sync_type = AV_SYNC_AUDIO_MASTER; static int64_t start_time = AV_NOPTS_VALUE; @@ -3105,7 +3106,13 @@ static VideoState *stream_open(const char *filename, AVInputFormat *iformat) init_clock(&is->audclk, &is->audioq.serial); init_clock(&is->extclk, &is->extclk.serial); is->audio_clock_serial = -1; - is->audio_volume = SDL_MIX_MAXVOLUME; + if (startup_volume < 0) + av_log(NULL, AV_LOG_WARNING, "-volume=%d < 0, setting to 0\n", startup_volume); + if (startup_volume > 100) + av_log(NULL, AV_LOG_WARNING, "-volume=%d > 100, setting to 100\n", startup_volume); + startup_volume = av_clip(startup_volume, 0, 100); + startup_volume = av_clip(SDL_MIX_MAXVOLUME * startup_volume / 100, 0, SDL_MIX_MAXVOLUME); + is->audio_volume = startup_volume; is->muted = 0; is->av_sync_type = av_sync_type; is->read_tid = SDL_CreateThread(read_thread, "read_thread", is); @@ -3586,6 +3593,7 @@ static const OptionDef options[] = { { "t", HAS_ARG, { .func_arg = opt_duration }, "play \"duration\" seconds of audio/video", "duration" }, { "bytes", OPT_INT | HAS_ARG, { &seek_by_bytes }, "seek by bytes 0=off 1=on -1=auto", "val" }, { "nodisp", OPT_BOOL, { &display_disable }, "disable graphical display" }, + { "volume", OPT_INT | HAS_ARG, { &startup_volume}, "set startup volume 0=min 100=max", "volume" }, { "f", HAS_ARG, { .func_arg = opt_format }, "force format", "fmt" }, { "pix_fmt", HAS_ARG | OPT_EXPERT | OPT_VIDEO, { .func_arg = opt_frame_pix_fmt }, "set pixel format", "format" }, { "stats", OPT_BOOL | OPT_EXPERT, { &show_status }, "show status", "" },

1 0

avfilter/af_amerge: properly handle unknown input layouts
by git＠videolan.org 25 Dec '16

25 Dec '16

ffmpeg | branch: master | Marton Balint <cus(a)passwd.hu> | Thu Dec 15 03:28:20 2016 +0100| [0db48ee4257c16f583b7b077fdbd13cfd0b9f037] | committer: Marton Balint avfilter/af_amerge: properly handle unknown input layouts Reviewed-by: Nicolas George <george(a)nsup.org> Signed-off-by: Marton Balint <cus(a)passwd.hu> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=0db48ee4257c16f58… --- libavfilter/af_amerge.c | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-) diff --git a/libavfilter/af_amerge.c b/libavfilter/af_amerge.c index 40bf7ab..8ea01e2 100644 --- a/libavfilter/af_amerge.c +++ b/libavfilter/af_amerge.c @@ -96,10 +96,15 @@ static int query_formats(AVFilterContext *ctx) av_get_channel_layout_string(buf, sizeof(buf), 0, inlayout[i]); av_log(ctx, AV_LOG_INFO, "Using \"%s\" for input %d\n", buf, i + 1); } - s->in[i].nb_ch = av_get_channel_layout_nb_channels(inlayout[i]); - if (outlayout & inlayout[i]) + s->in[i].nb_ch = FF_LAYOUT2COUNT(inlayout[i]); + if (s->in[i].nb_ch) { overlap++; - outlayout |= inlayout[i]; + } else { + s->in[i].nb_ch = av_get_channel_layout_nb_channels(inlayout[i]); + if (outlayout & inlayout[i]) + overlap++; + outlayout |= inlayout[i]; + } nb_ch += s->in[i].nb_ch; } if (nb_ch > SWR_CH_MAX) {

1 0

avcodec/error_resilience: Merge surrounding status checks
by git＠videolan.org 25 Dec '16

25 Dec '16

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Sun Dec 25 20:22:06 2016 +0100| [d5ecffbac69543185a2f6d91414dbe097645f62c] | committer: Michael Niedermayer avcodec/error_resilience: Merge surrounding status checks Simplifies code and is also faster Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=d5ecffbac69543185… --- libavcodec/error_resilience.c | 32 ++++++++++++-------------------- 1 file changed, 12 insertions(+), 20 deletions(-) diff --git a/libavcodec/error_resilience.c b/libavcodec/error_resilience.c index 79ac75a..741a9c8 100644 --- a/libavcodec/error_resilience.c +++ b/libavcodec/error_resilience.c @@ -376,7 +376,7 @@ static void v_block_filter(ERContext *s, uint8_t *dst, int w, int h, static void guess_mv(ERContext *s) { uint8_t *fixed = s->er_temp_buffer; -#define MV_FROZEN 3 +#define MV_FROZEN 4 #define MV_CHANGED 2 #define MV_UNCHANGED 1 const int mb_stride = s->mb_stride; @@ -467,27 +467,19 @@ static void guess_mv(ERContext *s) av_assert1(s->last_pic.f && s->last_pic.f->data[0]); j = 0; - if (mb_x > 0 && fixed[mb_xy - 1] == MV_FROZEN) - j = 1; - if (mb_x + 1 < mb_width && fixed[mb_xy + 1] == MV_FROZEN) - j = 1; - if (mb_y > 0 && fixed[mb_xy - mb_stride] == MV_FROZEN) - j = 1; - if (mb_y + 1 < mb_height && fixed[mb_xy + mb_stride] == MV_FROZEN) - j = 1; - if (j == 0) + if (mb_x > 0) + j |= fixed[mb_xy - 1]; + if (mb_x + 1 < mb_width) + j |= fixed[mb_xy + 1]; + if (mb_y > 0) + j |= fixed[mb_xy - mb_stride]; + if (mb_y + 1 < mb_height) + j |= fixed[mb_xy + mb_stride]; + + if (!(j & MV_FROZEN)) continue; - j = 0; - if (mb_x > 0 && fixed[mb_xy - 1 ] == MV_CHANGED) - j = 1; - if (mb_x + 1 < mb_width && fixed[mb_xy + 1 ] == MV_CHANGED) - j = 1; - if (mb_y > 0 && fixed[mb_xy - mb_stride] == MV_CHANGED) - j = 1; - if (mb_y + 1 < mb_height && fixed[mb_xy + mb_stride] == MV_CHANGED) - j = 1; - if (j == 0 && pass > 1) + if (!(j & MV_CHANGED) && pass > 1) continue; none_left = 0;

1 0

avcodec/error_resilience: Factor block parity check out
by git＠videolan.org 25 Dec '16

25 Dec '16

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Sun Dec 25 20:16:00 2016 +0100| [cafc72bd7bc4fb7672717e0e599405dbc59f3b5d] | committer: Michael Niedermayer avcodec/error_resilience: Factor block parity check out This makes the code noticably faster when there are lots of blocks Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=cafc72bd7bc4fb767… --- libavcodec/error_resilience.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) diff --git a/libavcodec/error_resilience.c b/libavcodec/error_resilience.c index 2790599..79ac75a 100644 --- a/libavcodec/error_resilience.c +++ b/libavcodec/error_resilience.c @@ -450,7 +450,7 @@ static void guess_mv(ERContext *s) changed = 0; for (mb_y = 0; mb_y < mb_height; mb_y++) { - for (mb_x = 0; mb_x < s->mb_width; mb_x++) { + for (mb_x = (mb_y ^ pass) & 1; mb_x < s->mb_width; mb_x+=2) { const int mb_xy = mb_x + mb_y * s->mb_stride; int mv_predictor[8][2] = { { 0 } }; int ref[8] = { 0 }; @@ -461,9 +461,6 @@ static void guess_mv(ERContext *s) const int mot_index = (mb_x + mb_y * mot_stride) * mot_step; int prev_x = 0, prev_y = 0, prev_ref = 0; - if ((mb_x ^ mb_y ^ pass) & 1) - continue; - if (fixed[mb_xy] == MV_FROZEN) continue; av_assert1(!IS_INTRA(s->cur_pic.mb_type[mb_xy]));

1 0

avfilter/vf_w3fdif: add >8 but <16 bit support
by git＠videolan.org 25 Dec '16

25 Dec '16

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Sun Sep 11 12:11:00 2016 +0200| [c6c888e996126b7189d27f16ba2aed38f8cff680] | committer: Paul B Mahol avfilter/vf_w3fdif: add >8 but <16 bit support Signed-off-by: Paul B Mahol <onemda(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=c6c888e996126b718… --- libavfilter/vf_w3fdif.c | 134 ++++++++++++++++++++++++++++++++++++--- libavfilter/w3fdif.h | 5 +- libavfilter/x86/vf_w3fdif_init.c | 9 +-- 3 files changed, 133 insertions(+), 15 deletions(-) diff --git a/libavfilter/vf_w3fdif.c b/libavfilter/vf_w3fdif.c index cde17d2..b7872db 100644 --- a/libavfilter/vf_w3fdif.c +++ b/libavfilter/vf_w3fdif.c @@ -43,6 +43,7 @@ typedef struct W3FDIFContext { AVFrame *prev, *cur, *next; ///< previous, current, next frames int32_t **work_line; ///< lines we are calculating int nb_threads; + int max; W3FDIFDSPContext dsp; } W3FDIFContext; @@ -75,6 +76,11 @@ static int query_formats(AVFilterContext *ctx) AV_PIX_FMT_YUVA420P, AV_PIX_FMT_YUVA422P, AV_PIX_FMT_YUVA444P, AV_PIX_FMT_GBRP, AV_PIX_FMT_GBRAP, AV_PIX_FMT_GRAY8, + AV_PIX_FMT_YUV420P9, AV_PIX_FMT_YUV422P9, AV_PIX_FMT_YUV444P9, + AV_PIX_FMT_YUV420P10, AV_PIX_FMT_YUV422P10, AV_PIX_FMT_YUV444P10, + AV_PIX_FMT_YUV420P12, AV_PIX_FMT_YUV422P12, AV_PIX_FMT_YUV444P12, + AV_PIX_FMT_YUV420P14, AV_PIX_FMT_YUV422P14, AV_PIX_FMT_YUV444P14, + AV_PIX_FMT_GBRP9, AV_PIX_FMT_GBRP10, AV_PIX_FMT_GBRP12, AV_PIX_FMT_GBRP14, AV_PIX_FMT_NONE }; @@ -148,7 +154,7 @@ static void filter_complex_high(int32_t *work_line, } } -static void filter_scale(uint8_t *out_pixel, const int32_t *work_pixel, int linesize) +static void filter_scale(uint8_t *out_pixel, const int32_t *work_pixel, int linesize, int max) { int j; @@ -156,12 +162,111 @@ static void filter_scale(uint8_t *out_pixel, const int32_t *work_pixel, int line *out_pixel = av_clip(*work_pixel, 0, 255 * 256 * 128) >> 15; } +static void filter16_simple_low(int32_t *work_line, + uint8_t *in_lines_cur8[2], + const int16_t *coef, int linesize) +{ + uint16_t *in_lines_cur[2] = { (uint16_t *)in_lines_cur8[0], (uint16_t *)in_lines_cur8[1] }; + int i; + + linesize /= 2; + for (i = 0; i < linesize; i++) { + *work_line = *in_lines_cur[0]++ * coef[0]; + *work_line++ += *in_lines_cur[1]++ * coef[1]; + } +} + +static void filter16_complex_low(int32_t *work_line, + uint8_t *in_lines_cur8[4], + const int16_t *coef, int linesize) +{ + uint16_t *in_lines_cur[4] = { (uint16_t *)in_lines_cur8[0], + (uint16_t *)in_lines_cur8[1], + (uint16_t *)in_lines_cur8[2], + (uint16_t *)in_lines_cur8[3] }; + int i; + + linesize /= 2; + for (i = 0; i < linesize; i++) { + *work_line = *in_lines_cur[0]++ * coef[0]; + *work_line += *in_lines_cur[1]++ * coef[1]; + *work_line += *in_lines_cur[2]++ * coef[2]; + *work_line++ += *in_lines_cur[3]++ * coef[3]; + } +} + +static void filter16_simple_high(int32_t *work_line, + uint8_t *in_lines_cur8[3], + uint8_t *in_lines_adj8[3], + const int16_t *coef, int linesize) +{ + uint16_t *in_lines_cur[3] = { (uint16_t *)in_lines_cur8[0], + (uint16_t *)in_lines_cur8[1], + (uint16_t *)in_lines_cur8[2] }; + uint16_t *in_lines_adj[3] = { (uint16_t *)in_lines_adj8[0], + (uint16_t *)in_lines_adj8[1], + (uint16_t *)in_lines_adj8[2] }; + int i; + + linesize /= 2; + for (i = 0; i < linesize; i++) { + *work_line += *in_lines_cur[0]++ * coef[0]; + *work_line += *in_lines_adj[0]++ * coef[0]; + *work_line += *in_lines_cur[1]++ * coef[1]; + *work_line += *in_lines_adj[1]++ * coef[1]; + *work_line += *in_lines_cur[2]++ * coef[2]; + *work_line++ += *in_lines_adj[2]++ * coef[2]; + } +} + +static void filter16_complex_high(int32_t *work_line, + uint8_t *in_lines_cur8[5], + uint8_t *in_lines_adj8[5], + const int16_t *coef, int linesize) +{ + uint16_t *in_lines_cur[5] = { (uint16_t *)in_lines_cur8[0], + (uint16_t *)in_lines_cur8[1], + (uint16_t *)in_lines_cur8[2], + (uint16_t *)in_lines_cur8[3], + (uint16_t *)in_lines_cur8[4] }; + uint16_t *in_lines_adj[5] = { (uint16_t *)in_lines_adj8[0], + (uint16_t *)in_lines_adj8[1], + (uint16_t *)in_lines_adj8[2], + (uint16_t *)in_lines_adj8[3], + (uint16_t *)in_lines_adj8[4] }; + int i; + + linesize /= 2; + for (i = 0; i < linesize; i++) { + *work_line += *in_lines_cur[0]++ * coef[0]; + *work_line += *in_lines_adj[0]++ * coef[0]; + *work_line += *in_lines_cur[1]++ * coef[1]; + *work_line += *in_lines_adj[1]++ * coef[1]; + *work_line += *in_lines_cur[2]++ * coef[2]; + *work_line += *in_lines_adj[2]++ * coef[2]; + *work_line += *in_lines_cur[3]++ * coef[3]; + *work_line += *in_lines_adj[3]++ * coef[3]; + *work_line += *in_lines_cur[4]++ * coef[4]; + *work_line++ += *in_lines_adj[4]++ * coef[4]; + } +} + +static void filter16_scale(uint8_t *out_pixel8, const int32_t *work_pixel, int linesize, int max) +{ + uint16_t *out_pixel = (uint16_t *)out_pixel8; + int j; + + linesize /= 2; + for (j = 0; j < linesize; j++, out_pixel++, work_pixel++) + *out_pixel = av_clip(*work_pixel, 0, max) >> 15; +} + static int config_input(AVFilterLink *inlink) { AVFilterContext *ctx = inlink->dst; W3FDIFContext *s = ctx->priv; const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format); - int ret, i; + int ret, i, depth; if ((ret = av_image_fill_linesizes(s->linesize, inlink->format, inlink->w)) < 0) return ret; @@ -181,14 +286,24 @@ static int config_input(AVFilterLink *inlink) return AVERROR(ENOMEM); } - s->dsp.filter_simple_low = filter_simple_low; - s->dsp.filter_complex_low = filter_complex_low; - s->dsp.filter_simple_high = filter_simple_high; - s->dsp.filter_complex_high = filter_complex_high; - s->dsp.filter_scale = filter_scale; + depth = desc->comp[0].depth; + s->max = ((1 << depth) - 1) * 256 * 128; + if (depth <= 8) { + s->dsp.filter_simple_low = filter_simple_low; + s->dsp.filter_complex_low = filter_complex_low; + s->dsp.filter_simple_high = filter_simple_high; + s->dsp.filter_complex_high = filter_complex_high; + s->dsp.filter_scale = filter_scale; + } else { + s->dsp.filter_simple_low = filter16_simple_low; + s->dsp.filter_complex_low = filter16_complex_low; + s->dsp.filter_simple_high = filter16_simple_high; + s->dsp.filter_complex_high = filter16_complex_high; + s->dsp.filter_scale = filter16_scale; + } if (ARCH_X86) - ff_w3fdif_init_x86(&s->dsp); + ff_w3fdif_init_x86(&s->dsp, depth); return 0; } @@ -247,6 +362,7 @@ static int deinterlace_slice(AVFilterContext *ctx, void *arg, int jobnr, int nb_ const int dst_line_stride = out->linesize[plane]; const int start = (height * jobnr) / nb_jobs; const int end = (height * (jobnr+1)) / nb_jobs; + const int max = s->max; int j, y_in, y_out; /* copy unchanged the lines of the field */ @@ -319,7 +435,7 @@ static int deinterlace_slice(AVFilterContext *ctx, void *arg, int jobnr, int nb_ work_pixel = s->work_line[jobnr]; out_pixel = out_line; - s->dsp.filter_scale(out_pixel, work_pixel, linesize); + s->dsp.filter_scale(out_pixel, work_pixel, linesize, max); /* move on to next line */ y_out += 2; diff --git a/libavfilter/w3fdif.h b/libavfilter/w3fdif.h index 67bb874..9c0b723 100644 --- a/libavfilter/w3fdif.h +++ b/libavfilter/w3fdif.h @@ -39,9 +39,10 @@ typedef struct W3FDIFDSPContext { uint8_t *in_lines_cur[5], uint8_t *in_lines_adj[5], const int16_t *coef, int linesize); - void (*filter_scale)(uint8_t *out_pixel, const int32_t *work_pixel, int linesize); + void (*filter_scale)(uint8_t *out_pixel, const int32_t *work_pixel, + int linesize, int max); } W3FDIFDSPContext; -void ff_w3fdif_init_x86(W3FDIFDSPContext *dsp); +void ff_w3fdif_init_x86(W3FDIFDSPContext *dsp, int depth); #endif /* AVFILTER_W3FDIF_H */ diff --git a/libavfilter/x86/vf_w3fdif_init.c b/libavfilter/x86/vf_w3fdif_init.c index 9bf06e8..d4534bb 100644 --- a/libavfilter/x86/vf_w3fdif_init.c +++ b/libavfilter/x86/vf_w3fdif_init.c @@ -43,20 +43,21 @@ void ff_w3fdif_complex_high_sse2(int32_t *work_line, uint8_t *in_lines_adj[5], const int16_t *coef, int linesize); -void ff_w3fdif_scale_sse2(uint8_t *out_pixel, const int32_t *work_pixel, int linesize); +void ff_w3fdif_scale_sse2(uint8_t *out_pixel, const int32_t *work_pixel, + int linesize, int max); -av_cold void ff_w3fdif_init_x86(W3FDIFDSPContext *dsp) +av_cold void ff_w3fdif_init_x86(W3FDIFDSPContext *dsp, int depth) { int cpu_flags = av_get_cpu_flags(); - if (EXTERNAL_SSE2(cpu_flags)) { + if (EXTERNAL_SSE2(cpu_flags) && depth <= 8) { dsp->filter_simple_low = ff_w3fdif_simple_low_sse2; dsp->filter_simple_high = ff_w3fdif_simple_high_sse2; dsp->filter_complex_low = ff_w3fdif_complex_low_sse2; dsp->filter_scale = ff_w3fdif_scale_sse2; } - if (ARCH_X86_64 && EXTERNAL_SSE2(cpu_flags)) { + if (ARCH_X86_64 && EXTERNAL_SSE2(cpu_flags) && depth <= 8) { dsp->filter_complex_high = ff_w3fdif_complex_high_sse2; } }

1 0