April 2020 - ffmpeg-devel - lists.ffmpeg.org

[PATCH 1/7] tools/target_dec_fuzzer: Fuzz skip_frame
by Michael Niedermayer 17 Jul '21

17 Jul '21

Should allow coverage of related code Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> --- tools/target_dec_fuzzer.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/tools/target_dec_fuzzer.c b/tools/target_dec_fuzzer.c index 2487b6ca94..67a2a41874 100644 --- a/tools/target_dec_fuzzer.c +++ b/tools/target_dec_fuzzer.c @@ -223,6 +223,8 @@ int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) { ctx->request_channel_layout = bytestream2_get_le64(&gbc); ctx->idct_algo = bytestream2_get_byte(&gbc) % 25; + ctx->skip_frame = bytestream2_get_byte(&gbc) - 254 + AVDISCARD_ALL; + if (flags & 0x20) { switch (ctx->codec_id) { -- 2.17.1

3 14

[PATCH] set AVFrame decode_error_flags in case of decoding error by h264dec
by Amir Pauker 31 May '21

31 May '21

set AVFrame decode_error_flags in case h->slice_ctx->er.error_occurred is set after the call to ff_h264_execute_decode_slices. This allows the user to detect concealed decoding errors in the call to avcodec_receive_frame Signed-off-by: Amir Pauker <amir(a)livelyvideo.tv> --- libavcodec/error_resilience.c | 2 ++ libavcodec/h264dec.c | 5 +++++ 2 files changed, 7 insertions(+) diff --git a/libavcodec/error_resilience.c b/libavcodec/error_resilience.c index 35d0c60..ca22871 100644 --- a/libavcodec/error_resilience.c +++ b/libavcodec/error_resilience.c @@ -1121,6 +1121,8 @@ void ff_er_frame_end(ERContext *s) av_log(s->avctx, AV_LOG_INFO, "concealing %d DC, %d AC, %d MV errors in %c frame\n", dc_error, ac_error, mv_error, av_get_picture_type_char(s->cur_pic.f->pict_type)); + s->cur_pic.f->decode_error_flags |= FF_DECODE_ERROR_CONCEALMENT_ACTIVE; + is_intra_likely = is_intra_more_likely(s); /* set unknown mb-type to most likely */ diff --git a/libavcodec/h264dec.c b/libavcodec/h264dec.c index 837c3b7..8d1bd16 100644 --- a/libavcodec/h264dec.c +++ b/libavcodec/h264dec.c @@ -761,6 +761,11 @@ static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size) if (ret < 0 && (h->avctx->err_recognition & AV_EF_EXPLODE)) goto end; + // set decode_error_flags to allow users to detect concealed decoding errors + if ((ret < 0 || h->slice_ctx->er.error_occurred) && h->cur_pic_ptr) { + h->cur_pic_ptr->f->decode_error_flags |= FF_DECODE_ERROR_DECODE_SLICES; + } + ret = 0; end: -- 2.1.4

5 5

[PATCH 1/2] examples/avio_dir_cmd: drop support for move/delete operations
by Anton Khirnov 22 May '21

22 May '21

They use non-public functions, which is unacceptable for a public API example. Rename the example back to avio_list_dir. This effectively reverts c84d208c275d6a43b3c3421d38772179abf8acee and 767d780ec001167b2fd8f6cfe4ef78a3a8b1e34c. --- configure | 4 +- doc/examples/.gitignore | 2 +- doc/examples/Makefile | 2 +- doc/examples/Makefile.example | 2 +- .../{avio_dir_cmd.c => avio_list_dir.c} | 56 ++----------------- 5 files changed, 9 insertions(+), 57 deletions(-) rename doc/examples/{avio_dir_cmd.c => avio_list_dir.c} (69%) diff --git a/configure b/configure index 46f2038627..65a5a30f36 100755 --- a/configure +++ b/configure @@ -1664,7 +1664,7 @@ COMPONENT_LIST=" " EXAMPLE_LIST=" - avio_dir_cmd_example + avio_list_dir_example avio_reading_example decode_audio_example decode_video_example @@ -3599,7 +3599,7 @@ yadif_cuda_filter_deps="ffnvcodec" yadif_cuda_filter_deps_any="cuda_nvcc cuda_llvm" # examples -avio_dir_cmd_deps="avformat avutil" +avio_list_dir_deps="avformat avutil" avio_reading_deps="avformat avcodec avutil" decode_audio_example_deps="avcodec avutil" decode_video_example_deps="avcodec avutil" diff --git a/doc/examples/.gitignore b/doc/examples/.gitignore index 75152cb50b..44960e1de7 100644 --- a/doc/examples/.gitignore +++ b/doc/examples/.gitignore @@ -1,4 +1,4 @@ -/avio_dir_cmd +/avio_list_dir /avio_reading /decode_audio /decode_video diff --git a/doc/examples/Makefile b/doc/examples/Makefile index 2935424e54..81bfd34d5d 100644 --- a/doc/examples/Makefile +++ b/doc/examples/Makefile @@ -1,4 +1,4 @@ -EXAMPLES-$(CONFIG_AVIO_DIR_CMD_EXAMPLE) += avio_dir_cmd +EXAMPLES-$(CONFIG_AVIO_LIST_DIR_EXAMPLE) += avio_list_dir EXAMPLES-$(CONFIG_AVIO_READING_EXAMPLE) += avio_reading EXAMPLES-$(CONFIG_DECODE_AUDIO_EXAMPLE) += decode_audio EXAMPLES-$(CONFIG_DECODE_VIDEO_EXAMPLE) += decode_video diff --git a/doc/examples/Makefile.example b/doc/examples/Makefile.example index 6428154c51..a232d97f98 100644 --- a/doc/examples/Makefile.example +++ b/doc/examples/Makefile.example @@ -11,7 +11,7 @@ CFLAGS += -Wall -g CFLAGS := $(shell pkg-config --cflags $(FFMPEG_LIBS)) $(CFLAGS) LDLIBS := $(shell pkg-config --libs $(FFMPEG_LIBS)) $(LDLIBS) -EXAMPLES= avio_dir_cmd \ +EXAMPLES= avio_list_dir \ avio_reading \ decode_audio \ decode_video \ diff --git a/doc/examples/avio_dir_cmd.c b/doc/examples/avio_list_dir.c similarity index 69% rename from doc/examples/avio_dir_cmd.c rename to doc/examples/avio_list_dir.c index 0722bd9ab1..3073baaefa 100644 --- a/doc/examples/avio_dir_cmd.c +++ b/doc/examples/avio_list_dir.c @@ -102,38 +102,15 @@ static int list_op(const char *input_dir) return ret; } -static int del_op(const char *url) -{ - int ret = avpriv_io_delete(url); - if (ret < 0) - av_log(NULL, AV_LOG_ERROR, "Cannot delete '%s': %s.\n", url, av_err2str(ret)); - return ret; -} - -static int move_op(const char *src, const char *dst) -{ - int ret = avpriv_io_move(src, dst); - if (ret < 0) - av_log(NULL, AV_LOG_ERROR, "Cannot move '%s' into '%s': %s.\n", src, dst, av_err2str(ret)); - return ret; -} - - static void usage(const char *program_name) { - fprintf(stderr, "usage: %s OPERATION entry1 [entry2]\n" - "API example program to show how to manipulate resources " - "accessed through AVIOContext.\n" - "OPERATIONS:\n" - "list list content of the directory\n" - "move rename content in directory\n" - "del delete content in directory\n", - program_name); + fprintf(stderr, "usage: %s input_dir\n" + "API example program to show how to list files in directory " + "accessed through AVIOContext.\n", program_name); } int main(int argc, char *argv[]) { - const char *op = NULL; int ret; av_log_set_level(AV_LOG_DEBUG); @@ -145,32 +122,7 @@ int main(int argc, char *argv[]) avformat_network_init(); - op = argv[1]; - if (strcmp(op, "list") == 0) { - if (argc < 3) { - av_log(NULL, AV_LOG_INFO, "Missing argument for list operation.\n"); - ret = AVERROR(EINVAL); - } else { - ret = list_op(argv[2]); - } - } else if (strcmp(op, "del") == 0) { - if (argc < 3) { - av_log(NULL, AV_LOG_INFO, "Missing argument for del operation.\n"); - ret = AVERROR(EINVAL); - } else { - ret = del_op(argv[2]); - } - } else if (strcmp(op, "move") == 0) { - if (argc < 4) { - av_log(NULL, AV_LOG_INFO, "Missing argument for move operation.\n"); - ret = AVERROR(EINVAL); - } else { - ret = move_op(argv[2], argv[3]); - } - } else { - av_log(NULL, AV_LOG_INFO, "Invalid operation %s\n", op); - ret = AVERROR(EINVAL); - } + ret = list_op(argv[1]); avformat_network_deinit(); -- 2.24.1

4 11

[PATCH 1/4] avcodec/startcode: Use AV_RN due to UBSan warning about unaligned access
by Andreas Rheinhardt 28 Mar '21

28 Mar '21

when directly accessing a buffer via a pointer to uint8_t cast to a pointer to uint64_t/uint32_t. This happened only if HAVE_FAST_UNALIGNED was set, so it was ok, but UBSan nevertheless complained about unaligned accesses. So simply use AV_RNxx to read the buffer; this also improves readability. This affected many FATE-tests (namely 228 of them; e.g. the cbs-h264 tests) and it has also been reported in tickets #8138 and #8485. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt(a)gmail.com> --- libavcodec/startcode.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/libavcodec/startcode.c b/libavcodec/startcode.c index 9efdffe8c6..d84f326521 100644 --- a/libavcodec/startcode.c +++ b/libavcodec/startcode.c @@ -25,6 +25,7 @@ * @author Michael Niedermayer <michaelni(a)gmx.at> */ +#include "libavutil/intreadwrite.h" #include "startcode.h" #include "config.h" @@ -38,14 +39,14 @@ int ff_startcode_find_candidate_c(const uint8_t *buf, int size) */ #if HAVE_FAST_64BIT while (i < size && - !((~*(const uint64_t *)(buf + i) & - (*(const uint64_t *)(buf + i) - 0x0101010101010101ULL)) & + !((~AV_RN64(buf + i) & + (AV_RN64(buf + i) - 0x0101010101010101ULL)) & 0x8080808080808080ULL)) i += 8; #else while (i < size && - !((~*(const uint32_t *)(buf + i) & - (*(const uint32_t *)(buf + i) - 0x01010101U)) & + !((~AV_RN32(buf + i) & + (AV_RN32(buf + i) - 0x01010101U)) & 0x80808080U)) i += 4; #endif -- 2.20.1

5 16

[PATCH v2 01/15] avfilter/vf_xbr: Fix left shift of negative number
by Andreas Rheinhardt 26 Mar '21

26 Mar '21

Affected every usage of vf_xbr, e.g. the FATE-tests filter-2xbr, filter-3xbr, filter-4xbr. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt(a)gmail.com> --- libavfilter/vf_xbr.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libavfilter/vf_xbr.c b/libavfilter/vf_xbr.c index 2c71871d22..87810c88f1 100644 --- a/libavfilter/vf_xbr.c +++ b/libavfilter/vf_xbr.c @@ -395,7 +395,7 @@ static int init(AVFilterContext *ctx) int startg = FFMAX3(-bg, -rg, 0); int endg = FFMIN3(255-bg, 255-rg, 255); uint32_t y = (uint32_t)(( 299*rg + 1000*startg + 114*bg)/1000); - c = bg + (rg<<16) + 0x010101 * startg; + c = bg + rg * (1 << 16) + 0x010101 * startg; for (g = startg; g <= endg; g++) { s->rgbtoyuv[c] = ((y++) << 16) + (u << 8) + v; c+= 0x010101; -- 2.20.1

4 28

[PATCH] avformat/matroskadec: Add webm file extension
by Andreas Rheinhardt 26 Feb '21

26 Feb '21

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt(a)gmail.com> --- libavformat/matroskadec.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libavformat/matroskadec.c b/libavformat/matroskadec.c index 75f72d330c..3879a54af5 100644 --- a/libavformat/matroskadec.c +++ b/libavformat/matroskadec.c @@ -4193,7 +4193,7 @@ static const AVClass webm_dash_class = { AVInputFormat ff_matroska_demuxer = { .name = "matroska,webm", .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"), - .extensions = "mkv,mk3d,mka,mks", + .extensions = "mkv,mk3d,mka,mks,webm", .priv_data_size = sizeof(MatroskaDemuxContext), .read_probe = matroska_probe, .read_header = matroska_read_header, -- 2.20.1

1 1

[PATCH v3] QP tables API
by Anton Khirnov 24 Feb '21

24 Feb '21

Hi, a new iteration of the QP tables/video encoding parameters patches. The set now passes FATE and could be pushed if there are no further comments. Changes from last iteration include: - type is now passed to the constructor, allowing for type-specific per-block data to be added later - QP type is now int32_t - TYPE_H263 (equivalent to FF_QSCALE_TYPE_MPEG1) is gone, since it's semantically equivalent to TYPE_MPEG2 divided by two. So we just multiply by two on export and do not clutter the API with too many types. 4/9 on its own breaks FATE (since the decoders start exporting the data in a different format), which is fixed by 5/9 and 6/9. We might want to merge those so that every commit passes FATE. Alternatively we could consider a compatibility layer, but given the obscurity of this feature that does not seem to be worth the effort. Some filters (codecview, fspp, pp7, spp, uspp), and mjpegdec, are not converted yet. Since those have no tests, this does not break FATE. I may convert them later or someone else is welcome to do it. Please comment -- Anton Khirnov

8 26

[PATCH 01/11] avformat/matroskadec: Reject sipr flavor > 3
by Andreas Rheinhardt 15 Feb '21

15 Feb '21

Only flavors 0..3 seem to exist. E.g. rmdec.c treats any flavor > 3 as invalid data. Furthermore, we do not know how big the packets to create ought to be given that for sipr these values are not read from the bitstream, but from a table. Furthermore, flavor is only used for sipr, so only check it for sipr; rmdec.c does the same. (The old check for flavor being < 0 was always wrong given that flavor is an int that is read via avio_rb16(), so it has been removed completely.) Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt(a)gmail.com> --- libavformat/matroskadec.c | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) diff --git a/libavformat/matroskadec.c b/libavformat/matroskadec.c index 8e1326abf6..8c65e98e77 100644 --- a/libavformat/matroskadec.c +++ b/libavformat/matroskadec.c @@ -2606,28 +2606,30 @@ static int matroska_parse_tracks(AVFormatContext *s) track->audio.sub_packet_h = avio_rb16(&b); track->audio.frame_size = avio_rb16(&b); track->audio.sub_packet_size = avio_rb16(&b); - if (flavor < 0 || - track->audio.coded_framesize <= 0 || + if (track->audio.coded_framesize <= 0 || track->audio.sub_packet_h <= 0 || track->audio.frame_size <= 0 || track->audio.sub_packet_size <= 0 && codec_id != AV_CODEC_ID_SIPR) return AVERROR_INVALIDDATA; - track->audio.buf = av_malloc_array(track->audio.sub_packet_h, - track->audio.frame_size); - if (!track->audio.buf) - return AVERROR(ENOMEM); + if (codec_id == AV_CODEC_ID_RA_288) { st->codecpar->block_align = track->audio.coded_framesize; track->codec_priv.size = 0; } else { - if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) { + if (codec_id == AV_CODEC_ID_SIPR) { static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 }; + if (flavor > 3) + return AVERROR_INVALIDDATA; track->audio.sub_packet_size = ff_sipr_subpk_size[flavor]; st->codecpar->bit_rate = sipr_bit_rate[flavor]; } st->codecpar->block_align = track->audio.sub_packet_size; extradata_offset = 78; } + track->audio.buf = av_malloc_array(track->audio.sub_packet_h, + track->audio.frame_size); + if (!track->audio.buf) + return AVERROR(ENOMEM); } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) { ret = matroska_parse_flac(s, track, &extradata_offset); if (ret < 0) -- 2.20.1

1 13

[PATCH] lavd/dshow: Add namespace prefix to global functions
by Mark Thompson 21 Jan '21

21 Jan '21

--- Not tested at all. libavdevice/dshow.c | 6 +++--- libavdevice/dshow_capture.h | 4 ++-- libavdevice/dshow_crossbar.c | 8 ++++---- 3 files changed, 9 insertions(+), 9 deletions(-) diff --git a/libavdevice/dshow.c b/libavdevice/dshow.c index de910c0994..6f613f31b0 100644 --- a/libavdevice/dshow.c +++ b/libavdevice/dshow.c @@ -509,7 +509,7 @@ end: * Pops up a user dialog allowing them to adjust properties for the given filter, if possible. */ void -dshow_show_filter_properties(IBaseFilter *device_filter, AVFormatContext *avctx) { +ff_dshow_show_filter_properties(IBaseFilter *device_filter, AVFormatContext *avctx) { ISpecifyPropertyPages *property_pages = NULL; IUnknown *device_filter_iunknown = NULL; HRESULT hr; @@ -581,7 +581,7 @@ dshow_cycle_pins(AVFormatContext *avctx, enum dshowDeviceType devtype, int should_show_properties = (devtype == VideoDevice) ? ctx->show_video_device_dialog : ctx->show_audio_device_dialog; if (should_show_properties) - dshow_show_filter_properties(device_filter, avctx); + ff_dshow_show_filter_properties(device_filter, avctx); r = IBaseFilter_EnumPins(device_filter, &pins); if (r != S_OK) { @@ -886,7 +886,7 @@ dshow_open_device(AVFormatContext *avctx, ICreateDevEnum *devenum, goto error; } - r = dshow_try_setup_crossbar_options(graph_builder2, device_filter, devtype, avctx); + r = ff_dshow_try_setup_crossbar_options(graph_builder2, device_filter, devtype, avctx); if (r != S_OK) { av_log(avctx, AV_LOG_ERROR, "Could not setup CrossBar\n"); diff --git a/libavdevice/dshow_capture.h b/libavdevice/dshow_capture.h index 475d62ba99..79c004a87f 100644 --- a/libavdevice/dshow_capture.h +++ b/libavdevice/dshow_capture.h @@ -346,9 +346,9 @@ struct dshow_ctx { /***************************************************************************** * CrossBar ****************************************************************************/ -HRESULT dshow_try_setup_crossbar_options(ICaptureGraphBuilder2 *graph_builder2, +HRESULT ff_dshow_try_setup_crossbar_options(ICaptureGraphBuilder2 *graph_builder2, IBaseFilter *device_filter, enum dshowDeviceType devtype, AVFormatContext *avctx); -void dshow_show_filter_properties(IBaseFilter *pFilter, AVFormatContext *avctx); +void ff_dshow_show_filter_properties(IBaseFilter *pFilter, AVFormatContext *avctx); #endif /* AVDEVICE_DSHOW_CAPTURE_H */ diff --git a/libavdevice/dshow_crossbar.c b/libavdevice/dshow_crossbar.c index 95fb466f4e..2438683cde 100644 --- a/libavdevice/dshow_crossbar.c +++ b/libavdevice/dshow_crossbar.c @@ -137,7 +137,7 @@ setup_crossbar_options(IAMCrossbar *cross_bar, enum dshowDeviceType devtype, AVF * Given a fully constructed graph, check if there is a cross bar filter, and configure its pins if so. */ HRESULT -dshow_try_setup_crossbar_options(ICaptureGraphBuilder2 *graph_builder2, +ff_dshow_try_setup_crossbar_options(ICaptureGraphBuilder2 *graph_builder2, IBaseFilter *device_filter, enum dshowDeviceType devtype, AVFormatContext *avctx) { struct dshow_ctx *ctx = avctx->priv_data; @@ -163,7 +163,7 @@ dshow_try_setup_crossbar_options(ICaptureGraphBuilder2 *graph_builder2, hr = IAMCrossbar_QueryInterface(cross_bar, &IID_IBaseFilter, (void **) &cross_bar_base_filter); if (hr != S_OK) goto end; - dshow_show_filter_properties(cross_bar_base_filter, avctx); + ff_dshow_show_filter_properties(cross_bar_base_filter, avctx); } if (devtype == VideoDevice && ctx->show_analog_tv_tuner_dialog) { @@ -173,7 +173,7 @@ dshow_try_setup_crossbar_options(ICaptureGraphBuilder2 *graph_builder2, hr = IAMCrossbar_QueryInterface(tv_tuner_filter, &IID_IBaseFilter, (void **) &tv_tuner_base_filter); if (hr != S_OK) goto end; - dshow_show_filter_properties(tv_tuner_base_filter, avctx); + ff_dshow_show_filter_properties(tv_tuner_base_filter, avctx); } else { av_log(avctx, AV_LOG_WARNING, "unable to find a tv tuner to display dialog for!"); } @@ -185,7 +185,7 @@ dshow_try_setup_crossbar_options(ICaptureGraphBuilder2 *graph_builder2, hr = IAMCrossbar_QueryInterface(tv_audio_filter, &IID_IBaseFilter, (void **) &tv_audio_base_filter); if (hr != S_OK) goto end; - dshow_show_filter_properties(tv_audio_base_filter, avctx); + ff_dshow_show_filter_properties(tv_audio_base_filter, avctx); } else { av_log(avctx, AV_LOG_WARNING, "unable to find a tv audio tuner to display dialog for!"); } -- 2.18.0

2 2

[RFC PATCH] avfilter/fastdeint: import simple cpu-optimized deinterlacing algorithms from VLC
by Aman Gupta 02 Jan '21

02 Jan '21

From: Aman Gupta <aman(a)tmm1.net> These are simple algorithms which can be run efficiently on low powered devices to produce deinteraced images. Signed-off-by: Aman Gupta <aman(a)tmm1.net> --- doc/filters.texi | 27 ++ libavfilter/Makefile | 1 + libavfilter/aarch64/Makefile | 1 + libavfilter/aarch64/merge_neon.S | 98 ++++++ libavfilter/allfilters.c | 1 + libavfilter/arm/Makefile | 3 + libavfilter/arm/merge_armv6.S | 70 ++++ libavfilter/arm/merge_neon.S | 109 ++++++ libavfilter/vf_fastdeint.c | 588 +++++++++++++++++++++++++++++++ 9 files changed, 898 insertions(+) create mode 100644 libavfilter/aarch64/merge_neon.S create mode 100644 libavfilter/arm/Makefile create mode 100644 libavfilter/arm/merge_armv6.S create mode 100644 libavfilter/arm/merge_neon.S create mode 100644 libavfilter/vf_fastdeint.c diff --git a/doc/filters.texi b/doc/filters.texi index 6c81e1da40..55d9adeb81 100644 --- a/doc/filters.texi +++ b/doc/filters.texi @@ -9796,6 +9796,33 @@ fade=t=in:st=5.5:d=0.5 @end itemize +@section fastdeint +Fast deinterlacing algorithms. + +@table @option +@item mode +Deinterlacing algorithm to use. + +It accepts the following values: +@table @samp +@item discard +Discard bottom frame. + +@item mean +Half resolution blender. + +@item blend +Full resolution blender. + +@item bob +Bob doubler. + +@item linear +Bob doubler with linear interpolation. +@end table + +@end table + @section fftdnoiz Denoise frames using 3D FFT (frequency domain filtering). diff --git a/libavfilter/Makefile b/libavfilter/Makefile index 3ef4191d9a..a2b3566ec0 100644 --- a/libavfilter/Makefile +++ b/libavfilter/Makefile @@ -234,6 +234,7 @@ OBJS-$(CONFIG_EROSION_OPENCL_FILTER) += vf_neighbor_opencl.o opencl.o \ opencl/neighbor.o OBJS-$(CONFIG_EXTRACTPLANES_FILTER) += vf_extractplanes.o OBJS-$(CONFIG_FADE_FILTER) += vf_fade.o +OBJS-$(CONFIG_FASTDEINT_FILTER) += vf_fastdeint.o OBJS-$(CONFIG_FFTDNOIZ_FILTER) += vf_fftdnoiz.o OBJS-$(CONFIG_FFTFILT_FILTER) += vf_fftfilt.o OBJS-$(CONFIG_FIELD_FILTER) += vf_field.o diff --git a/libavfilter/aarch64/Makefile b/libavfilter/aarch64/Makefile index b58daa3a3f..2b0ad92893 100644 --- a/libavfilter/aarch64/Makefile +++ b/libavfilter/aarch64/Makefile @@ -1,3 +1,4 @@ OBJS-$(CONFIG_NLMEANS_FILTER) += aarch64/vf_nlmeans_init.o +NEON-OBJS-$(CONFIG_FASTDEINT_FILTER) += aarch64/merge_neon.o NEON-OBJS-$(CONFIG_NLMEANS_FILTER) += aarch64/vf_nlmeans_neon.o diff --git a/libavfilter/aarch64/merge_neon.S b/libavfilter/aarch64/merge_neon.S new file mode 100644 index 0000000000..62377331a4 --- /dev/null +++ b/libavfilter/aarch64/merge_neon.S @@ -0,0 +1,98 @@ +/* + * Copyright (c) 2009-2016 Rémi Denis-Courmont, Janne Grunau, VLC authors + * + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "libavutil/aarch64/asm.S" + +#define dest x0 +#define src1 x1 +#define src2 x2 +#define size x3 + + .align 2 + // NOTE: Offset and pitch must be multiple of 16-bytes. +function ff_merge8_neon, export=1 + ands x5, size, #~63 + b.eq 2f + mov x10, #64 + add x11, src1, #32 + add x12, src2, #32 +1: + ld1 {v0.16b,v1.16b}, [src1], x10 + ld1 {v4.16b,v5.16b}, [src2], x10 + ld1 {v2.16b,v3.16b}, [x11], x10 + uhadd v0.16b, v0.16b, v4.16b + ld1 {v6.16b,v7.16b}, [x12], x10 + subs x5, x5, #64 + uhadd v1.16b, v1.16b, v5.16b + uhadd v2.16b, v2.16b, v6.16b + uhadd v3.16b, v3.16b, v7.16b + st1 {v0.16b,v1.16b}, [dest], #32 + st1 {v2.16b,v3.16b}, [dest], #32 + b.gt 1b +2: + tbz size, #5, 3f + ld1 {v0.16b,v1.16b}, [src1], #32 + ld1 {v4.16b,v5.16b}, [src2], #32 + uhadd v0.16b, v0.16b, v4.16b + uhadd v1.16b, v1.16b, v5.16b + st1 {v0.16b,v1.16b}, [dest], #32 +3: + tbz size, #4, 4f + ld1 {v0.16b}, [src1] + ld1 {v4.16b}, [src2] + uhadd v0.16b, v0.16b, v4.16b + st1 {v0.16b}, [dest] +4: + ret +endfunc + + .align 2 +function ff_merge16_neon, export=1 + ands x5, size, #~63 + b.eq 2f +1: + ld1 {v0.8h,v1.8h}, [src1], #32 + ld1 {v4.8h,v5.8h}, [src2], #32 + ld1 {v2.8h,v3.8h}, [src1], #32 + uhadd v0.8h, v0.8h, v4.8h + ld1 {v6.8h,v7.8h}, [src2], #32 + uhadd v1.8h, v1.8h, v5.8h + uhadd v2.8h, v2.8h, v6.8h + uhadd v3.8h, v3.8h, v7.8h + st1 {v0.8h,v1.8h}, [dest], #32 + st1 {v2.8h,v3.8h}, [dest], #32 + subs x5, x5, #64 + b.gt 1b +2: + tbz size, #5, 3f + ld1 {v0.8h,v1.8h}, [src1], #32 + ld1 {v4.8h,v5.8h}, [src2], #32 + uhadd v0.8h, v0.8h, v4.8h + uhadd v1.8h, v1.8h, v5.8h + st1 {v0.8h,v1.8h}, [dest], #32 +3: + tbz size, #4, 4f + ld1 {v0.8h}, [src1] + ld1 {v4.8h}, [src2] + uhadd v0.8h, v0.8h,v4.8h + st1 {v0.8h}, [dest] +4: + ret +endfunc diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c index b675c688ee..6631af2ffe 100644 --- a/libavfilter/allfilters.c +++ b/libavfilter/allfilters.c @@ -219,6 +219,7 @@ extern AVFilter ff_vf_erosion; extern AVFilter ff_vf_erosion_opencl; extern AVFilter ff_vf_extractplanes; extern AVFilter ff_vf_fade; +extern AVFilter ff_vf_fastdeint; extern AVFilter ff_vf_fftdnoiz; extern AVFilter ff_vf_fftfilt; extern AVFilter ff_vf_field; diff --git a/libavfilter/arm/Makefile b/libavfilter/arm/Makefile new file mode 100644 index 0000000000..c92d62fac9 --- /dev/null +++ b/libavfilter/arm/Makefile @@ -0,0 +1,3 @@ +ARMV6-OBJS-$(CONFIG_FASTDEINT_FILTER) += arm/merge_armv6.o + +NEON-OBJS-$(CONFIG_FASTDEINT_FILTER) += arm/merge_neon.o diff --git a/libavfilter/arm/merge_armv6.S b/libavfilter/arm/merge_armv6.S new file mode 100644 index 0000000000..9b551c2c6c --- /dev/null +++ b/libavfilter/arm/merge_armv6.S @@ -0,0 +1,70 @@ +/* + * Copyright (c) 2009-2012 Rémi Denis-Courmont, VLC authors + * + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "libavutil/arm/asm.S" + +#define dest r0 +#define src1 r1 +#define src2 r2 +#define size r3 + + .align 2 +function ff_merge8_armv6, export=1 + push {r4-r9,lr} +1: + pld [src1, #64] + ldm src1!, {r4-r5} + pld [src2, #64] + ldm src2!, {r8-r9} + subs size, size, #16 + uhadd8 r4, r4, r8 + ldm src1!, {r6-r7} + uhadd8 r5, r5, r9 + ldm src2!, {ip,lr} + uhadd8 r6, r6, ip + stm dest!, {r4-r5} + uhadd8 r7, r7, lr + stm dest!, {r6-r7} + it eq + popeq {r4-r9,pc} + b 1b +endfunc + + .align 2 +function ff_merge16_armv6, export=1 + push {r4-r9,lr} +1: + pld [src1, #64] + ldm src1!, {r4-r5} + pld [src2, #64] + ldm src2!, {r8-r9} + subs size, size, #16 + uhadd16 r4, r4, r8 + ldm src1!, {r6-r7} + uhadd16 r5, r5, r9 + ldm src2!, {ip,lr} + uhadd16 r6, r6, ip + stm dest!, {r4-r5} + uhadd16 r7, r7, lr + stm dest!, {r6-r7} + it eq + popeq {r4-r9,pc} + b 1b +endfunc \ No newline at end of file diff --git a/libavfilter/arm/merge_neon.S b/libavfilter/arm/merge_neon.S new file mode 100644 index 0000000000..ae36cf3ca9 --- /dev/null +++ b/libavfilter/arm/merge_neon.S @@ -0,0 +1,109 @@ +/* + * Copyright (c) 2009-2012 Rémi Denis-Courmont, VLC authors + * + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "libavutil/arm/asm.S" + +#define dest r0 +#define src1 r1 +#define src2 r2 +#define size r3 + + .align 2 + @ NOTE: Offset and pitch must be multiple of 16-bytes. +function ff_merge8_neon, export=1 + cmp size, #64 + blo 2f +1: + pld [src1, #64] + vld1.u8 {q0-q1}, [src1,:128]! + pld [src2, #64] + vld1.u8 {q8-q9}, [src2,:128]! + vhadd.u8 q0, q0, q8 + sub size, size, #64 + vld1.u8 {q2-q3}, [src1,:128]! + vhadd.u8 q1, q1, q9 + vld1.u8 {q10-q11}, [src2,:128]! + vhadd.u8 q2, q2, q10 + cmp size, #64 + vhadd.u8 q3, q3, q11 + vst1.u8 {q0-q1}, [dest,:128]! + vst1.u8 {q2-q3}, [dest,:128]! + bhs 1b +2: + cmp size, #32 + blo 3f + vld1.u8 {q0-q1}, [src1,:128]! + sub size, size, #32 + vld1.u8 {q8-q9}, [src2,:128]! + vhadd.u8 q0, q0, q8 + vhadd.u8 q1, q1, q9 + vst1.u8 {q0-q1}, [dest,:128]! +3: + cmp size, #16 + it lo + bxlo lr + vld1.u8 {q0}, [src1,:128]! + sub size, size, #16 + vld1.u8 {q8}, [src2,:128]! + vhadd.u8 q0, q0, q8 + vst1.u8 {q0}, [dest,:128]! + bx lr +endfunc + + .align 2 +function ff_merge16_neon, export=1 + cmp size, #64 + blo 2f +1: + pld [src1, #64] + vld1.u16 {q0-q1}, [src1,:128]! + pld [src2, #64] + vld1.u16 {q8-q9}, [src2,:128]! + vhadd.u16 q0, q0, q8 + sub size, size, #64 + vld1.u16 {q2-q3}, [src1,:128]! + vhadd.u16 q1, q1, q9 + vld1.u16 {q10-q11}, [src2,:128]! + vhadd.u16 q2, q2, q10 + cmp size, #64 + vhadd.u16 q3, q3, q11 + vst1.u16 {q0-q1}, [dest,:128]! + vst1.u16 {q2-q3}, [dest,:128]! + bhs 1b +2: + cmp size, #32 + blo 3f + vld1.u16 {q0-q1}, [src1,:128]! + sub size, size, #32 + vld1.u16 {q8-q9}, [src2,:128]! + vhadd.u16 q0, q0, q8 + vhadd.u16 q1, q1, q9 + vst1.u16 {q0-q1}, [dest,:128]! +3: + cmp size, #16 + it lo + bxlo lr + vld1.u16 {q0}, [src1,:128]! + sub size, size, #16 + vld1.u16 {q8}, [src2,:128]! + vhadd.u16 q0, q0, q8 + vst1.u16 {q0}, [dest,:128]! + bx lr +endfunc \ No newline at end of file diff --git a/libavfilter/vf_fastdeint.c b/libavfilter/vf_fastdeint.c new file mode 100644 index 0000000000..5ddd8be392 --- /dev/null +++ b/libavfilter/vf_fastdeint.c @@ -0,0 +1,588 @@ +/* + * Copyright (C) 2015 Aman Gupta <aman(a)tmm1.net> + * 2000-2011 VLC authors and VideoLAN + * + * Author: Sam Hocevar <sam(a)zoy.org> + * Damien Lucas <nitrox(a)videolan.org> + * Laurent Aimar <fenrir(a)videolan.org> + * Sigmund Augdal Helberg <sigmunau(a)videolan.org> + * + * These algorithms are derived from the VLC project's + * modules/video_filter/deinterlace/algo_basic.c + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "libavutil/avassert.h" +#include "libavutil/cpu.h" +#include "libavutil/common.h" +#include "libavutil/opt.h" +#include "libavutil/pixdesc.h" +#include "libavutil/imgutils.h" +#include "libavutil/timestamp.h" +#include "avfilter.h" +#include "formats.h" +#include "internal.h" +#include "video.h" + +enum Mode { + MODE_DISCARD, + MODE_MEAN, + MODE_BLEND, + MODE_BOB, + MODE_LINEAR, + MODE_MAX, +}; + +typedef void (*merge_fn)(void *dst, const void *src1, const void *src2, size_t len); + +typedef struct FastDeintContext { + const AVClass *class; + merge_fn merge; + int merge_size; + int merge_aligned; + AVFrame *cur, *next; + enum Mode mode; + int eof; +} FastDeintContext; + +static void merge8_c(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, size_t bytes) +{ + for (; bytes > 0; bytes--) + *dst++ = ( *src1++ + *src2++ ) >> 1; +} + +static void merge16_c(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, size_t bytes) +{ + for (size_t words = bytes / 2; words > 0; words--) + *dst++ = ( *src1++ + *src2++ ) >> 1; +} + +static void merge8_unaligned(FastDeintContext *s, uint8_t *dst, const uint8_t *src1, const uint8_t *src2, size_t bytes) +{ + if (s->merge_aligned) { + size_t remainder = bytes % 16; + if (remainder > 0) { + merge8_c(dst, src1, src2, remainder); + bytes -= remainder; + dst += remainder; + src1 += remainder; + src2 += remainder; + } + } + s->merge(dst, src1, src2, bytes); +} + +static void merge16_unaligned(FastDeintContext *s, uint16_t *dst, const uint16_t *src1, const uint16_t *src2, size_t bytes) +{ + if (s->merge_aligned) { + size_t words = bytes / 2; + size_t remainder = words % 8; + if (remainder > 0) { + merge16_c(dst, src1, src2, remainder); + words -= remainder; + dst += remainder; + src1 += remainder; + src2 += remainder; + } + } + s->merge(dst, src1, src2, bytes); +} + +static void merge_unaligned(FastDeintContext *s, void *dst, const void *src1, const void *src2, size_t bytes) +{ + if (s->merge_size == 16) + merge16_unaligned(s, dst, src1, src2, bytes); + else + merge8_unaligned(s, dst, src1, src2, bytes); +} + +#if HAVE_SSE2_INLINE && defined(__x86_64__) +static void merge8_sse2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, size_t bytes) +{ + for(; bytes > 0 && ((uintptr_t)src1 & 15); bytes--) + *dst++ = ( *src1++ + *src2++ ) >> 1; + + for (; bytes >= 16; bytes -= 16) { + __asm__ __volatile__( "movdqu %2,%%xmm1;" + "pavgb %1, %%xmm1;" + "movdqu %%xmm1, %0" :"=m" (*dst): + "m" (*src1), + "m" (*src2) : "xmm1" ); + dst += 16; + src1 += 16; + src2 += 16; + } + + if (bytes > 0) { + merge8_c(dst, src1, src2, bytes); + } +} +static void merge16_sse2(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, size_t bytes) +{ + size_t words = bytes / 2; + + for(; words > 0 && ((uintptr_t)src1 & 15); words--) + *dst++ = ( *src1++ + *src2++ ) >> 1; + + for (; words >= 8; words -= 8) { + __asm__ __volatile__( "movdqu %2,%%xmm1;" + "pavgw %1, %%xmm1;" + "movdqu %%xmm1, %0" :"=m" (*dst): + "m" (*src1), + "m" (*src2) : "xmm1" ); + dst += 8; + src1 += 8; + src2 += 8; + } + + if (words > 0) { + merge16_c(dst, src1, src2, words * 2); + } +} +#define merge8 merge8_sse2 +#define merge16 merge16_sse2 +#else +#define merge8 merge8_c +#define merge16 merge16_c +#endif + +static void render_image_single(FastDeintContext *s, AVFrame *out, AVFrame *frame) +{ + int i, planes_nb = 0; + enum Mode mode = s->mode; + const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(out->format); + + for (i = 0; i < desc->nb_components; i++) + planes_nb = FFMAX(planes_nb, desc->comp[i].plane + 1); + + for (i = 0; i < planes_nb; i++) { + int height, bwidth; + int dst_linesize, src_linesize; + const uint8_t *src; + uint8_t *dst; + + bwidth = av_image_get_linesize(out->format, out->width, i); + if (bwidth < 0) { + av_log(s, AV_LOG_ERROR, "av_image_get_linesize failed\n"); + return; + } + + height = out->height; + if (i == 1 || i == 2) { + height = FF_CEIL_RSHIFT(out->height, desc->log2_chroma_h); + } + + src = frame->data[i]; + dst = out->data[i]; + dst_linesize = out->linesize[i]; + src_linesize = frame->linesize[i]; + + if (mode == MODE_BLEND) { + // Copy first line + memcpy(dst, src, bwidth); + dst += dst_linesize; + height--; + } + + // Merge remaining lines + for (; height > 0; height--) { + if (mode == MODE_DISCARD) + memcpy(dst, src, bwidth); + else + merge_unaligned(s, dst, src, src + src_linesize, bwidth); + dst += dst_linesize; + src += src_linesize; + if (mode == MODE_MEAN || mode == MODE_DISCARD) { + src += src_linesize; + height--; + } + } + } + if (mode != MODE_DISCARD) + emms_c(); +} + +static void render_image_doubler(FastDeintContext *s, AVFrame *out, AVFrame *frame, int field) +{ + int i, planes_nb = 0; + enum Mode mode = s->mode; + const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(out->format); + + for (i = 0; i < desc->nb_components; i++) + planes_nb = FFMAX(planes_nb, desc->comp[i].plane + 1); + + for (i = 0; i < planes_nb; i++) { + int height, bwidth; + int dst_linesize, src_linesize; + const uint8_t *src; + uint8_t *dst; + + bwidth = av_image_get_linesize(out->format, out->width, i); + if (bwidth < 0) { + av_log(s, AV_LOG_ERROR, "av_image_get_linesize failed\n"); + return; + } + height = out->height; + if (i == 1 || i == 2) { + height = FF_CEIL_RSHIFT(out->height, desc->log2_chroma_h); + } + + src = frame->data[i]; + dst = out->data[i]; + src_linesize = frame->linesize[i]; + dst_linesize = out->linesize[i]; + + // For BOTTOM field we need to add the first line + if (field == 1) { + memcpy(dst, src, bwidth); + dst += dst_linesize; + src += src_linesize; + height--; + } + + height -= 2; + + for (; height > 0; height-=2) { + memcpy(dst, src, bwidth); + dst += dst_linesize; + + if (mode == MODE_LINEAR) + merge_unaligned(s, dst, src, src + 2 * src_linesize, bwidth); + else + memcpy(dst, src, bwidth); + dst += dst_linesize; + + src += src_linesize * 2; + } + + memcpy(dst, src, bwidth); + + // For TOP field we need to add the last line + if (field == 0) + { + dst += dst_linesize; + src += src_linesize; + memcpy(dst, src, bwidth); + } + } + if (mode == MODE_LINEAR) + emms_c(); +} + +static int filter_frame_single(AVFilterLink *link, AVFrame *frame) +{ + AVFilterContext *ctx = link->dst; + AVFrame *out; + FastDeintContext *s = ctx->priv; + + if (!frame->interlaced_frame) { + return ff_filter_frame(ctx->outputs[0], frame); + } + + out = ff_get_video_buffer(ctx->outputs[0], link->w, link->h); + if (!out) { + av_frame_free(&frame); + return AVERROR(ENOMEM); + } + + av_frame_copy_props(out, frame); + out->interlaced_frame = 0; + render_image_single(s, out, frame); + + av_frame_free(&frame); + return ff_filter_frame(ctx->outputs[0], out); +} + +static AVFrame *copy_frame(AVFilterLink *link, AVFrame *frame) +{ + AVFilterContext *ctx = link->dst; + AVFrame *out; + + if (frame->format == AV_PIX_FMT_VIDEOTOOLBOX) + out = av_frame_alloc(); + else + out = ff_get_video_buffer(ctx->outputs[0], link->w, link->h); + + if (!out) + return NULL; + + av_frame_copy_props(out, frame); + return out; +} + +static int filter_frame_double(AVFilterLink *link, AVFrame *in) +{ + AVFilterContext *ctx = link->dst; + FastDeintContext *s = ctx->priv; + AVFrame *frame, *out, *out2; + int tff, ret; + + s->cur = s->next; + s->next = in; + + if (!s->cur) { + return 0; + } + + frame = s->cur; + + if (!frame->interlaced_frame) { + if (frame->pts != AV_NOPTS_VALUE) + frame->pts *= 2; + s->cur = NULL; + return ff_filter_frame(ctx->outputs[0], frame); + } + + tff = frame->top_field_first; + out = copy_frame(link, frame); + if (!out) { + av_frame_free(&frame); + s->cur = NULL; + return AVERROR(ENOMEM); + } + + out->interlaced_frame = 0; + if (out->pts != AV_NOPTS_VALUE) + out->pts = out->pts * 2; + render_image_doubler(s, out, frame, !tff); + + ret = ff_filter_frame(ctx->outputs[0], out); + if (ret < 0) { + av_frame_free(&frame); + s->cur = NULL; + return ret; + } + + out2 = copy_frame(link, frame); + if (!out2) { + av_frame_free(&frame); + s->cur = NULL; + return AVERROR(ENOMEM); + } + + out2->interlaced_frame = 0; + av_frame_remove_side_data(out2, AV_FRAME_DATA_A53_CC); + if (out2->pts != AV_NOPTS_VALUE) { + out2->pts = frame->pts + s->next->pts; + } + render_image_doubler(s, out2, frame, tff); + + av_frame_free(&frame); + s->cur = NULL; + + return ff_filter_frame(ctx->outputs[0], out2); +} + +static int filter_frame(AVFilterLink *link, AVFrame *frame) +{ + AVFilterContext *ctx = link->dst; + FastDeintContext *s = ctx->priv; + + av_assert0(frame); + + if (s->mode == MODE_LINEAR || s->mode == MODE_BOB) { + return filter_frame_double(link, frame); + } else { + return filter_frame_single(link, frame); + } +} + +static av_cold void uninit(AVFilterContext *ctx) +{ + FastDeintContext *s = ctx->priv; + av_frame_free(&s->cur); + av_frame_free(&s->next); +} + +static int query_formats(AVFilterContext *ctx) +{ + static const enum AVPixelFormat pix_fmts[] = { + AV_PIX_FMT_YUV420P, + AV_PIX_FMT_YUV422P, + AV_PIX_FMT_YUV444P, + AV_PIX_FMT_YUV410P, + AV_PIX_FMT_YUV411P, + AV_PIX_FMT_GRAY8, + AV_PIX_FMT_YUVJ420P, + AV_PIX_FMT_YUVJ422P, + AV_PIX_FMT_YUVJ444P, + AV_PIX_FMT_GRAY16, + AV_PIX_FMT_YUV440P, + AV_PIX_FMT_YUVJ440P, + 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_YUV420P16, + AV_PIX_FMT_YUV422P16, + AV_PIX_FMT_YUV444P16, + AV_PIX_FMT_YUVA420P, + AV_PIX_FMT_YUVA422P, + AV_PIX_FMT_YUVA444P, + AV_PIX_FMT_GBRP, + AV_PIX_FMT_GBRP9, + AV_PIX_FMT_GBRP10, + AV_PIX_FMT_GBRP12, + AV_PIX_FMT_GBRP14, + AV_PIX_FMT_GBRP16, + AV_PIX_FMT_GBRAP, + AV_PIX_FMT_NONE + }; + + AVFilterFormats *fmts_list = ff_make_format_list(pix_fmts); + if (!fmts_list) + return AVERROR(ENOMEM); + return ff_set_common_formats(ctx, fmts_list); +} + +#if ARCH_ARM +#include "libavutil/arm/cpu.h" +#endif +#if ARCH_AARCH64 +#include "libavutil/aarch64/cpu.h" +#endif +#if ARCH_AARCH64 || ARCH_ARM +void ff_merge8_neon(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, size_t bytes); +void ff_merge16_neon(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, size_t bytes); +void ff_merge8_armv6(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, size_t bytes); +void ff_merge16_armv6(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, size_t bytes); +#endif + +static int config_props(AVFilterLink *link) +{ + AVFilterContext *ctx = link->src; + FastDeintContext *s = ctx->priv; + const AVPixFmtDescriptor *pix; +#if ARCH_AARCH64 || ARCH_ARM + int cpu_flags = av_get_cpu_flags(); +#endif + + link->w = link->src->inputs[0]->w; + link->h = link->src->inputs[0]->h; + link->time_base = link->src->inputs[0]->time_base; + link->frame_rate = link->src->inputs[0]->frame_rate; + link->sample_aspect_ratio = link->src->inputs[0]->sample_aspect_ratio; + + if (s->mode == MODE_MEAN || s->mode == MODE_DISCARD) { + link->h /= 2; + link->sample_aspect_ratio = av_mul_q(link->sample_aspect_ratio, av_make_q(1, 2)); + } + if (s->mode == MODE_LINEAR || s->mode == MODE_BOB) { + link->time_base = av_mul_q(link->time_base, av_make_q(1, 2)); + link->frame_rate = av_mul_q(link->frame_rate, av_make_q(2, 1)); + } + + pix = av_pix_fmt_desc_get(link->format); + s->merge_size = (pix->comp[0].depth > 8) ? 16 : 8; + s->merge = s->merge_size == 16 ? (merge_fn)merge16 : (merge_fn)merge8; + +#if ARCH_ARM + if (have_armv6(cpu_flags)) { + s->merge = s->merge_size == 16 ? (merge_fn)ff_merge16_armv6 : (merge_fn)ff_merge8_armv6; + s->merge_aligned = 1; + } +#endif +#if ARCH_AARCH64 || ARCH_ARM + if (have_neon(cpu_flags)) { + s->merge = s->merge_size == 16 ? (merge_fn)ff_merge16_neon : (merge_fn)ff_merge8_neon; + s->merge_aligned = 1; + } +#endif + + return 0; +} + +static int request_frame(AVFilterLink *link) +{ + AVFilterContext *ctx = link->src; + FastDeintContext *s = ctx->priv; + int ret; + + if (s->eof) + return AVERROR_EOF; + + ret = ff_request_frame(ctx->inputs[0]); + + if (ret == AVERROR_EOF && s->cur) { + AVFrame *next = av_frame_clone(s->next); + if (!next) + return AVERROR(ENOMEM); + + next->pts = s->next->pts * 2 - s->cur->pts; + filter_frame(ctx->inputs[0], next); + s->eof = 1; + } else if (ret < 0) { + return ret; + } + + return 0; +} + +#define OFFSET(x) offsetof(FastDeintContext, x) +#define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM + +#define CONST(name, help, val, unit) { name, help, 0, AV_OPT_TYPE_CONST, {.i64=val}, INT_MIN, INT_MAX, FLAGS, unit } + +static const AVOption fastdeint_options[] = { + { "mode", "specify the deinterlacing mode", OFFSET(mode), AV_OPT_TYPE_INT, {.i64=MODE_BLEND}, 0, MODE_MAX-1, FLAGS, "mode" }, + CONST("discard", "discard bottom frame", MODE_DISCARD, "mode"), + CONST("mean", "half resolution blender", MODE_MEAN, "mode"), + CONST("blend", "full resolution blender", MODE_BLEND, "mode"), + CONST("bob", "bob doubler", MODE_BOB, "mode"), + CONST("linear", "bob doubler with linear interpolation", MODE_LINEAR, "mode"), + + { NULL } +}; + +AVFILTER_DEFINE_CLASS(fastdeint); + +static const AVFilterPad fastdeint_inputs[] = { + { + .name = "default", + .type = AVMEDIA_TYPE_VIDEO, + .filter_frame = filter_frame, + }, + { NULL } +}; + +static const AVFilterPad fastdeint_outputs[] = { + { + .name = "default", + .type = AVMEDIA_TYPE_VIDEO, + .config_props = config_props, + .request_frame = request_frame + }, + { NULL } +}; + +AVFilter ff_vf_fastdeint = { + .name = "fastdeint", + .description = NULL_IF_CONFIG_SMALL("fast deinterlacing algorithms"), + .priv_size = sizeof(FastDeintContext), + .priv_class = &fastdeint_class, + .uninit = uninit, + .query_formats = query_formats, + .inputs = fastdeint_inputs, + .outputs = fastdeint_outputs, +}; -- 2.20.1

5 8