January 2023 - ffmpeg-cvslog - lists.ffmpeg.org

doc/general_contents: sync with available decoders and demuxers
by Paul B Mahol 30 Jan '23

30 Jan '23

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Mon Jan 30 21:34:04 2023 +0100| [229f3259f80d824c7e7bb3bd5935ae4bc75d5374] | committer: Paul B Mahol doc/general_contents: sync with available decoders and demuxers > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=229f3259f80d824c7… --- doc/general_contents.texi | 13 +++++++++++++ 1 file changed, 13 insertions(+) diff --git a/doc/general_contents.texi b/doc/general_contents.texi index 8399fcb6b7..87e180c979 100644 --- a/doc/general_contents.texi +++ b/doc/general_contents.texi @@ -534,6 +534,8 @@ library: @item Metal Gear Solid: The Twin Snakes @tab @tab X @item Megalux Frame @tab @tab X @tab Used by Megalux Ultimate Paint +@item MobiClip MODS @tab @tab X +@item MobiClip MOFLEX @tab @tab X @item Mobotix .mxg @tab @tab X @item Monkey's Audio @tab @tab X @item Motion Pixels MVI @tab @tab X @@ -665,6 +667,7 @@ library: @item SBG @tab @tab X @item SDP @tab @tab X @item SER @tab @tab X +@item Digital Pictures SGA @tab @tab X @item Sega FILM/CPK @tab X @tab X @tab Used in many Sega Saturn console games. @item Silicon Graphics Movie @tab @tab X @@ -702,6 +705,7 @@ library: @item Vivo @tab @tab X @item VPK @tab @tab X @tab Audio format used in Sony PS games. +@item Marble WADY @tab @tab X @item WAV @tab X @tab X @item WavPack @tab X @tab X @item WebM @tab X @tab X @@ -714,6 +718,7 @@ library: @tab Multimedia format used in Westwood Studios games. @item Wideband Single-bit Data (WSD) @tab @tab X @item WVE @tab @tab X +@item Konami XMD @tab @tab X @item XMV @tab @tab X @tab Microsoft video container used in Xbox games. @item XVAG @tab @tab X @@ -763,6 +768,7 @@ following image formats are supported: @item JPEG-LS @tab X @tab X @item LJPEG @tab X @tab @tab Lossless JPEG +@item Media 100 @tab @tab X @item MSP @tab @tab X @tab Microsoft Paint image @item PAM @tab X @tab X @@ -1159,6 +1165,7 @@ following image formats are supported: @item ADPCM IMA Electronic Arts SEAD @tab @tab X @item ADPCM IMA Funcom @tab @tab X @item ADPCM IMA High Voltage Software ALP @tab X @tab X +@item ADPCM IMA Mobiclip MOFLEX @tab @tab X @item ADPCM IMA QuickTime @tab X @tab X @item ADPCM IMA Simon & Schuster Interactive @tab X @tab X @item ADPCM IMA Ubisoft APM @tab X @tab X @@ -1188,6 +1195,7 @@ following image formats are supported: @item ADPCM Sound Blaster Pro 4-bit @tab @tab X @item ADPCM VIMA @tab @tab X @tab Used in LucasArts SMUSH animations. +@item ADPCM Konami XMD @tab @tab X @item ADPCM Westwood Studios IMA @tab X @tab X @tab Used in Westwood Studios games like Command and Conquer. @item ADPCM Yamaha @tab X @tab X @@ -1225,9 +1233,12 @@ following image formats are supported: @item DCA (DTS Coherent Acoustics) @tab X @tab X @tab supported extensions: XCh, XXCH, X96, XBR, XLL, LBR (partially) @item Dolby E @tab @tab X +@item DPCM Cuberoot-Delta-Exact @tab @tab X + @tab Used in few games. @item DPCM Gremlin @tab @tab X @item DPCM id RoQ @tab X @tab X @tab Used in Quake III, Jedi Knight 2 and other computer games. +@item DPCM Marble WADY @tab @tab X @item DPCM Interplay @tab @tab X @tab Used in various Interplay computer games. @item DPCM Squareroot-Delta-Exact @tab @tab X @@ -1273,6 +1284,7 @@ following image formats are supported: @item MP3 (MPEG audio layer 3) @tab E @tab IX @tab encoding supported through external library LAME, ADU MP3 and MP3onMP4 also supported @item MPEG-4 Audio Lossless Coding (ALS) @tab @tab X +@item MobiClip FastAudio @tab @tab X @item Musepack SV7 @tab @tab X @item Musepack SV8 @tab @tab X @item Nellymoser Asao @tab X @tab X @@ -1307,6 +1319,7 @@ following image formats are supported: @item PCM unsigned 24-bit little-endian @tab X @tab X @item PCM unsigned 32-bit big-endian @tab X @tab X @item PCM unsigned 32-bit little-endian @tab X @tab X +@item PCM SGA @tab @tab X @item QCELP / PureVoice @tab @tab X @item QDesign Music Codec 1 @tab @tab X @item QDesign Music Codec 2 @tab @tab X

1 0

avformat: add XMD demuxer
by Paul B Mahol 30 Jan '23

30 Jan '23

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Wed Jan 25 19:16:03 2023 +0100| [fa42575ffa30f226f6fda89cb192f315319f8575] | committer: Paul B Mahol avformat: add XMD demuxer > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=fa42575ffa30f226f… --- Changelog | 2 +- libavformat/Makefile | 1 + libavformat/allformats.c | 1 + libavformat/version.h | 2 +- libavformat/xmd.c | 80 ++++++++++++++++++++++++++++++++++++++++++++++++ 5 files changed, 84 insertions(+), 2 deletions(-) diff --git a/Changelog b/Changelog index a0ba25257d..aa0dad13a1 100644 --- a/Changelog +++ b/Changelog @@ -34,7 +34,7 @@ version <next>: - ssim360 video filter - ffmpeg CLI new options: -enc_stats_pre[_fmt], -enc_stats_post[_fmt] - hstack_vaapi, vstack_vaapi and xstack_vaapi filters -- XMD ADPCM decoder +- XMD ADPCM decoder and demuxer version 5.1: diff --git a/libavformat/Makefile b/libavformat/Makefile index 2d11bcd7a3..0bb9e8c19a 100644 --- a/libavformat/Makefile +++ b/libavformat/Makefile @@ -621,6 +621,7 @@ OBJS-$(CONFIG_WVE_DEMUXER) += wvedec.o pcm.o OBJS-$(CONFIG_WV_MUXER) += wvenc.o wv.o apetag.o img2.o OBJS-$(CONFIG_XA_DEMUXER) += xa.o OBJS-$(CONFIG_XBIN_DEMUXER) += bintext.o sauce.o +OBJS-$(CONFIG_XMD_DEMUXER) += xmd.o pcm.o OBJS-$(CONFIG_XMV_DEMUXER) += xmv.o OBJS-$(CONFIG_XVAG_DEMUXER) += xvag.o OBJS-$(CONFIG_XWMA_DEMUXER) += xwma.o diff --git a/libavformat/allformats.c b/libavformat/allformats.c index bf8afe2078..2d31956fd5 100644 --- a/libavformat/allformats.c +++ b/libavformat/allformats.c @@ -501,6 +501,7 @@ extern const AVInputFormat ff_wv_demuxer; extern const AVOutputFormat ff_wv_muxer; extern const AVInputFormat ff_xa_demuxer; extern const AVInputFormat ff_xbin_demuxer; +extern const AVInputFormat ff_xmd_demuxer; extern const AVInputFormat ff_xmv_demuxer; extern const AVInputFormat ff_xvag_demuxer; extern const AVInputFormat ff_xwma_demuxer; diff --git a/libavformat/version.h b/libavformat/version.h index 3970ec5cdd..789859fbe2 100644 --- a/libavformat/version.h +++ b/libavformat/version.h @@ -31,7 +31,7 @@ #include "version_major.h" -#define LIBAVFORMAT_VERSION_MINOR 36 +#define LIBAVFORMAT_VERSION_MINOR 37 #define LIBAVFORMAT_VERSION_MICRO 100 #define LIBAVFORMAT_VERSION_INT AV_VERSION_INT(LIBAVFORMAT_VERSION_MAJOR, \ diff --git a/libavformat/xmd.c b/libavformat/xmd.c new file mode 100644 index 0000000000..f9c8138d72 --- /dev/null +++ b/libavformat/xmd.c @@ -0,0 +1,80 @@ +/* + * XMD demuxer + * Copyright (c) 2023 Paul B Mahol + * + * 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/channel_layout.h" +#include "libavutil/intreadwrite.h" +#include "avformat.h" +#include "demux.h" +#include "internal.h" +#include "pcm.h" + +static int xmd_probe(const AVProbeData *p) +{ + if ((AV_RL32(p->buf) & 0xFFFFFF) != MKTAG('x','m','d',0)) + return 0; + if (p->buf[3] == 0 || p->buf[3] > 2 || + AV_RL16(p->buf+4) == 0 || + AV_RL32(p->buf+6) == 0) + return 0; + + return AVPROBE_SCORE_MAX / 3; +} + +static int xmd_read_header(AVFormatContext *s) +{ + AVIOContext *pb = s->pb; + AVCodecParameters *par; + int channels; + AVStream *st; + + avio_skip(pb, 3); + + st = avformat_new_stream(s, NULL); + if (!st) + return AVERROR(ENOMEM); + + par = st->codecpar; + par->codec_type = AVMEDIA_TYPE_AUDIO; + par->codec_id = AV_CODEC_ID_ADPCM_XMD; + channels = avio_r8(pb); + if (channels == 0) + return AVERROR_INVALIDDATA; + av_channel_layout_default(&par->ch_layout, channels); + par->sample_rate = avio_rl16(pb); + if (par->sample_rate <= 0) + return AVERROR_INVALIDDATA; + par->block_align = 21 * channels; + st->duration = (avio_rl32(pb) / par->block_align) * 32LL; + avpriv_set_pts_info(st, 64, 1, par->sample_rate); + avio_skip(pb, 7); + + return 0; +} + +const AVInputFormat ff_xmd_demuxer = { + .name = "xmd", + .long_name = NULL_IF_CONFIG_SMALL("Konami XMD"), + .read_probe = xmd_probe, + .read_header = xmd_read_header, + .read_packet = ff_pcm_read_packet, + .flags = AVFMT_GENERIC_INDEX, + .extensions = "xmd", +};

1 0

avcodec: add ADPCM XMD decoder
by Paul B Mahol 30 Jan '23

30 Jan '23

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Wed Jan 25 19:31:44 2023 +0100| [280e6e93fc96afbbb73f1987eb3d603f8949ea30] | committer: Paul B Mahol avcodec: add ADPCM XMD decoder > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=280e6e93fc96afbbb… --- Changelog | 1 + libavcodec/Makefile | 1 + libavcodec/adpcm.c | 44 ++++++++++++++++++++++++++++++++++++++++++++ libavcodec/allcodecs.c | 1 + libavcodec/codec_desc.c | 7 +++++++ libavcodec/codec_id.h | 1 + libavcodec/utils.c | 3 +++ libavcodec/version.h | 2 +- 8 files changed, 59 insertions(+), 1 deletion(-) diff --git a/Changelog b/Changelog index a0f1ad7211..a0ba25257d 100644 --- a/Changelog +++ b/Changelog @@ -34,6 +34,7 @@ version <next>: - ssim360 video filter - ffmpeg CLI new options: -enc_stats_pre[_fmt], -enc_stats_post[_fmt] - hstack_vaapi, vstack_vaapi and xstack_vaapi filters +- XMD ADPCM decoder version 5.1: diff --git a/libavcodec/Makefile b/libavcodec/Makefile index f0ffd0b961..629933d85c 100644 --- a/libavcodec/Makefile +++ b/libavcodec/Makefile @@ -959,6 +959,7 @@ OBJS-$(CONFIG_ADPCM_THP_DECODER) += adpcm.o adpcm_data.o OBJS-$(CONFIG_ADPCM_THP_LE_DECODER) += adpcm.o adpcm_data.o OBJS-$(CONFIG_ADPCM_VIMA_DECODER) += vima.o adpcm_data.o OBJS-$(CONFIG_ADPCM_XA_DECODER) += adpcm.o adpcm_data.o +OBJS-$(CONFIG_ADPCM_XMD_DECODER) += adpcm.o adpcm_data.o OBJS-$(CONFIG_ADPCM_YAMAHA_DECODER) += adpcm.o adpcm_data.o OBJS-$(CONFIG_ADPCM_YAMAHA_ENCODER) += adpcmenc.o adpcm_data.o OBJS-$(CONFIG_ADPCM_ZORK_DECODER) += adpcm.o adpcm_data.o diff --git a/libavcodec/adpcm.c b/libavcodec/adpcm.c index 841538b138..451696932d 100644 --- a/libavcodec/adpcm.c +++ b/libavcodec/adpcm.c @@ -324,6 +324,7 @@ static av_cold int adpcm_decode_init(AVCodecContext * avctx) case AV_CODEC_ID_ADPCM_IMA_WAV: case AV_CODEC_ID_ADPCM_4XM: case AV_CODEC_ID_ADPCM_XA: + case AV_CODEC_ID_ADPCM_XMD: case AV_CODEC_ID_ADPCM_EA_R1: case AV_CODEC_ID_ADPCM_EA_R2: case AV_CODEC_ID_ADPCM_EA_R3: @@ -1043,6 +1044,9 @@ static int get_nb_samples(AVCodecContext *avctx, GetByteContext *gb, case AV_CODEC_ID_ADPCM_XA: nb_samples = (buf_size / 128) * 224 / ch; break; + case AV_CODEC_ID_ADPCM_XMD: + nb_samples = buf_size / (21 * ch) * 32; + break; case AV_CODEC_ID_ADPCM_DTK: case AV_CODEC_ID_ADPCM_PSX: nb_samples = buf_size / (16 * ch) * 28; @@ -1553,6 +1557,45 @@ static int adpcm_decode_frame(AVCodecContext *avctx, AVFrame *frame, } bytestream2_seek(&gb, 0, SEEK_END); ) /* End of CASE */ + CASE(ADPCM_XMD, + int bytes_remaining, block = 0; + while (bytestream2_get_bytes_left(&gb) >= 21 * channels) { + for (int channel = 0; channel < channels; channel++) { + int16_t *out = samples_p[channel] + block * 32; + int16_t history[2]; + uint16_t scale; + + history[1] = sign_extend(bytestream2_get_le16(&gb), 16); + history[0] = sign_extend(bytestream2_get_le16(&gb), 16); + scale = bytestream2_get_le16(&gb); + + out[0] = history[1]; + out[1] = history[0]; + + for (int n = 0; n < 15; n++) { + unsigned byte = bytestream2_get_byte(&gb); + int32_t nibble[2]; + + nibble[0] = sign_extend(byte & 15, 4); + nibble[1] = sign_extend(byte >> 4, 4); + + out[2+n*2] = (nibble[0]*(scale<<14) + (history[0]*29336) - (history[1]*13136)) >> 14; + history[1] = history[0]; + history[0] = out[2+n*2]; + + out[2+n*2+1] = (nibble[1]*(scale<<14) + (history[0]*29336) - (history[1]*13136)) >> 14; + history[1] = history[0]; + history[0] = out[2+n*2+1]; + } + } + + block++; + } + bytes_remaining = bytestream2_get_bytes_left(&gb); + if (bytes_remaining > 0) { + bytestream2_skip(&gb, bytes_remaining); + } + ) /* End of CASE */ CASE(ADPCM_XA, int16_t *out0 = samples_p[0]; int16_t *out1 = samples_p[1]; @@ -2350,5 +2393,6 @@ ADPCM_DECODER(ADPCM_SWF, sample_fmts_s16, adpcm_swf, "ADPCM Sho ADPCM_DECODER(ADPCM_THP_LE, sample_fmts_s16p, adpcm_thp_le, "ADPCM Nintendo THP (little-endian)") ADPCM_DECODER(ADPCM_THP, sample_fmts_s16p, adpcm_thp, "ADPCM Nintendo THP") ADPCM_DECODER(ADPCM_XA, sample_fmts_s16p, adpcm_xa, "ADPCM CDROM XA") +ADPCM_DECODER(ADPCM_XMD, sample_fmts_s16p, adpcm_xmd, "ADPCM Konami XMD") ADPCM_DECODER(ADPCM_YAMAHA, sample_fmts_s16, adpcm_yamaha, "ADPCM Yamaha") ADPCM_DECODER(ADPCM_ZORK, sample_fmts_s16, adpcm_zork, "ADPCM Zork") diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c index f30047e17a..0deaea65b3 100644 --- a/libavcodec/allcodecs.c +++ b/libavcodec/allcodecs.c @@ -695,6 +695,7 @@ extern const FFCodec ff_adpcm_thp_decoder; extern const FFCodec ff_adpcm_thp_le_decoder; extern const FFCodec ff_adpcm_vima_decoder; extern const FFCodec ff_adpcm_xa_decoder; +extern const FFCodec ff_adpcm_xmd_decoder; extern const FFCodec ff_adpcm_yamaha_encoder; extern const FFCodec ff_adpcm_yamaha_decoder; extern const FFCodec ff_adpcm_zork_decoder; diff --git a/libavcodec/codec_desc.c b/libavcodec/codec_desc.c index 50f9794e10..fbb6604f76 100644 --- a/libavcodec/codec_desc.c +++ b/libavcodec/codec_desc.c @@ -2536,6 +2536,13 @@ static const AVCodecDescriptor codec_descriptors[] = { .long_name = NULL_IF_CONFIG_SMALL("ADPCM IMA Acorn Replay"), .props = AV_CODEC_PROP_INTRA_ONLY | AV_CODEC_PROP_LOSSY, }, + { + .id = AV_CODEC_ID_ADPCM_XMD, + .type = AVMEDIA_TYPE_AUDIO, + .name = "adpcm_xmd", + .long_name = NULL_IF_CONFIG_SMALL("ADPCM Konami XMD"), + .props = AV_CODEC_PROP_INTRA_ONLY | AV_CODEC_PROP_LOSSY, + }, /* AMR */ { diff --git a/libavcodec/codec_id.h b/libavcodec/codec_id.h index cdf7eb79c3..3e84098f35 100644 --- a/libavcodec/codec_id.h +++ b/libavcodec/codec_id.h @@ -413,6 +413,7 @@ enum AVCodecID { AV_CODEC_ID_ADPCM_IMA_CUNNING, AV_CODEC_ID_ADPCM_IMA_MOFLEX, AV_CODEC_ID_ADPCM_IMA_ACORN, + AV_CODEC_ID_ADPCM_XMD, /* AMR */ AV_CODEC_ID_AMR_NB = 0x12000, diff --git a/libavcodec/utils.c b/libavcodec/utils.c index 94f7ae6877..18a433b1af 100644 --- a/libavcodec/utils.c +++ b/libavcodec/utils.c @@ -769,6 +769,9 @@ static int get_audio_frame_duration(enum AVCodecID id, int sr, int ch, int ba, case AV_CODEC_ID_ADPCM_MTAF: tmp = blocks * (ba - 16LL) * 2 / ch; break; + case AV_CODEC_ID_ADPCM_XMD: + tmp = blocks * 32; + break; } if (tmp) { if (tmp != (int)tmp) diff --git a/libavcodec/version.h b/libavcodec/version.h index 752adc81f8..2ed4ef5547 100644 --- a/libavcodec/version.h +++ b/libavcodec/version.h @@ -29,7 +29,7 @@ #include "version_major.h" -#define LIBAVCODEC_VERSION_MINOR 59 +#define LIBAVCODEC_VERSION_MINOR 60 #define LIBAVCODEC_VERSION_MICRO 100 #define LIBAVCODEC_VERSION_INT AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \

1 0

avfilter/vf_libplacebo: suppress cast warning
by Niklas Haas 30 Jan '23

30 Jan '23

ffmpeg | branch: master | Niklas Haas <git(a)haasn.dev> | Mon Jan 30 20:34:55 2023 +0100| [9bfdd8524e4d3657be4057798bd4f015322827b2] | committer: Niklas Haas avfilter/vf_libplacebo: suppress cast warning This warning was introduced when ee65039 removed the cast. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=9bfdd8524e4d3657b… --- libavfilter/vf_libplacebo.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/libavfilter/vf_libplacebo.c b/libavfilter/vf_libplacebo.c index d3c62d12a4..12abca4cc7 100644 --- a/libavfilter/vf_libplacebo.c +++ b/libavfilter/vf_libplacebo.c @@ -262,8 +262,7 @@ static int init_vulkan(AVFilterContext *avctx) return AVERROR(EINVAL); } - avhwctx = avctx->hw_device_ctx->data; - + avhwctx = (AVHWDeviceContext *) avctx->hw_device_ctx->data; if (avhwctx->type != AV_HWDEVICE_TYPE_VULKAN) { av_log(s, AV_LOG_ERROR, "Expected vulkan hwdevice for vf_libplacebo, got %s.\n", av_hwdevice_get_type_name(avhwctx->type));

1 0

avfilter/vf_colorlevels: add support for gbrpf32 format
by Paul B Mahol 30 Jan '23

30 Jan '23

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Mon Jan 30 17:01:26 2023 +0100| [5508e04c7b710616e33407ac2b3f8d616cb745f0] | committer: Paul B Mahol avfilter/vf_colorlevels: add support for gbrpf32 format > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=5508e04c7b710616e… --- libavfilter/vf_colorlevels.c | 118 +++++++++++++++++++++++++++++++++---------- 1 file changed, 90 insertions(+), 28 deletions(-) diff --git a/libavfilter/vf_colorlevels.c b/libavfilter/vf_colorlevels.c index 76adfbd8c8..44ae6cffbf 100644 --- a/libavfilter/vf_colorlevels.c +++ b/libavfilter/vf_colorlevels.c @@ -97,11 +97,13 @@ typedef struct ThreadData { int h; + float fimin[4]; + float fomin[4]; int imin[4]; int omin[4]; } ThreadData; -#define DO_COMMON(type, clip, preserve, planar) \ +#define DO_COMMON(type, ptype, clip, preserve, planar) \ const ThreadData *td = arg; \ const int linesize = s->linesize; \ const int step = s->step; \ @@ -118,14 +120,14 @@ typedef struct ThreadData { type *dst_g = (type *)(td->dstrow[G]) + src_linesize * slice_start; \ type *dst_b = (type *)(td->dstrow[B]) + src_linesize * slice_start; \ type *dst_a = (type *)(td->dstrow[A]) + src_linesize * slice_start; \ - const int imin_r = td->imin[R]; \ - const int imin_g = td->imin[G]; \ - const int imin_b = td->imin[B]; \ - const int imin_a = td->imin[A]; \ - const int omin_r = td->omin[R]; \ - const int omin_g = td->omin[G]; \ - const int omin_b = td->omin[B]; \ - const int omin_a = td->omin[A]; \ + const ptype imin_r = s->depth == 32 ? td->fimin[R] : td->imin[R]; \ + const ptype imin_g = s->depth == 32 ? td->fimin[G] : td->imin[G]; \ + const ptype imin_b = s->depth == 32 ? td->fimin[B] : td->imin[B]; \ + const ptype imin_a = s->depth == 32 ? td->fimin[A] : td->imin[A]; \ + const ptype omin_r = s->depth == 32 ? td->fomin[R] : td->omin[R]; \ + const ptype omin_g = s->depth == 32 ? td->fomin[G] : td->omin[G]; \ + const ptype omin_b = s->depth == 32 ? td->fomin[B] : td->omin[B]; \ + const ptype omin_a = s->depth == 32 ? td->fomin[A] : td->omin[A]; \ const float coeff_r = td->coeff[R]; \ const float coeff_g = td->coeff[G]; \ const float coeff_b = td->coeff[B]; \ @@ -133,12 +135,12 @@ typedef struct ThreadData { \ for (int y = slice_start; y < slice_end; y++) { \ for (int x = 0; x < linesize; x += step) { \ - int ir, ig, ib, or, og, ob; \ + ptype ir, ig, ib, or, og, ob; \ ir = src_r[x]; \ ig = src_g[x]; \ ib = src_b[x]; \ if (preserve) { \ - float ratio, icolor, ocolor, max = s->max; \ + float ratio, icolor, ocolor, max = s->depth==32 ? 1.f : s->max; \ \ or = (ir - imin_r) * coeff_r + omin_r; \ og = (ig - imin_g) * coeff_g + omin_g; \ @@ -180,39 +182,40 @@ typedef struct ThreadData { #define CLIP8(x, depth) av_clip_uint8(x) #define CLIP16(x, depth) av_clip_uint16(x) +#define NOCLIP(x, depth) (x) static int colorlevels_slice_8(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) { ColorLevelsContext *s = ctx->priv; - DO_COMMON(uint8_t, CLIP8, 0, 0) + DO_COMMON(uint8_t, int, CLIP8, 0, 0) return 0; } static int colorlevels_slice_16(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) { ColorLevelsContext *s = ctx->priv;\ - DO_COMMON(uint16_t, CLIP16, 0, 0) + DO_COMMON(uint16_t, int, CLIP16, 0, 0) return 0; } static int colorlevels_preserve_slice_8(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) { ColorLevelsContext *s = ctx->priv; - DO_COMMON(uint8_t, CLIP8, 1, 0) + DO_COMMON(uint8_t, int, CLIP8, 1, 0) return 0; } static int colorlevels_preserve_slice_16(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) { ColorLevelsContext *s = ctx->priv; - DO_COMMON(uint16_t, CLIP16, 1, 0) + DO_COMMON(uint16_t, int, CLIP16, 1, 0) return 0; } static int colorlevels_slice_8_planar(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) { ColorLevelsContext *s = ctx->priv; - DO_COMMON(uint8_t, CLIP8, 0, 1) + DO_COMMON(uint8_t, int, CLIP8, 0, 1) return 0; } @@ -220,7 +223,7 @@ static int colorlevels_slice_9_planar(AVFilterContext *ctx, void *arg, int jobnr { ColorLevelsContext *s = ctx->priv; const int depth = 9; - DO_COMMON(uint16_t, av_clip_uintp2, 0, 1) + DO_COMMON(uint16_t, int, av_clip_uintp2, 0, 1) return 0; } @@ -228,7 +231,7 @@ static int colorlevels_slice_10_planar(AVFilterContext *ctx, void *arg, int jobn { ColorLevelsContext *s = ctx->priv; const int depth = 10; - DO_COMMON(uint16_t, av_clip_uintp2, 0, 1) + DO_COMMON(uint16_t, int, av_clip_uintp2, 0, 1) return 0; } @@ -236,7 +239,7 @@ static int colorlevels_slice_12_planar(AVFilterContext *ctx, void *arg, int jobn { ColorLevelsContext *s = ctx->priv; const int depth = 12; - DO_COMMON(uint16_t, av_clip_uintp2, 0, 1) + DO_COMMON(uint16_t, int, av_clip_uintp2, 0, 1) return 0; } @@ -244,21 +247,28 @@ static int colorlevels_slice_14_planar(AVFilterContext *ctx, void *arg, int jobn { ColorLevelsContext *s = ctx->priv; const int depth = 14; - DO_COMMON(uint16_t, av_clip_uintp2, 0, 1) + DO_COMMON(uint16_t, int, av_clip_uintp2, 0, 1) return 0; } static int colorlevels_slice_16_planar(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) { ColorLevelsContext *s = ctx->priv; - DO_COMMON(uint16_t, CLIP16, 0, 1) + DO_COMMON(uint16_t, int, CLIP16, 0, 1) + return 0; +} + +static int colorlevels_slice_32_planar(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) +{ + ColorLevelsContext *s = ctx->priv; + DO_COMMON(float, float, NOCLIP, 0, 1) return 0; } static int colorlevels_preserve_slice_8_planar(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) { ColorLevelsContext *s = ctx->priv; - DO_COMMON(uint8_t, CLIP8, 1, 1) + DO_COMMON(uint8_t, int, CLIP8, 1, 1) return 0; } @@ -266,7 +276,7 @@ static int colorlevels_preserve_slice_9_planar(AVFilterContext *ctx, void *arg, { ColorLevelsContext *s = ctx->priv; const int depth = 9; - DO_COMMON(uint16_t, av_clip_uintp2, 1, 1) + DO_COMMON(uint16_t, int, av_clip_uintp2, 1, 1) return 0; } @@ -274,7 +284,7 @@ static int colorlevels_preserve_slice_10_planar(AVFilterContext *ctx, void *arg, { ColorLevelsContext *s = ctx->priv; const int depth = 10; - DO_COMMON(uint16_t, av_clip_uintp2, 1, 1) + DO_COMMON(uint16_t, int, av_clip_uintp2, 1, 1) return 0; } @@ -282,7 +292,7 @@ static int colorlevels_preserve_slice_12_planar(AVFilterContext *ctx, void *arg, { ColorLevelsContext *s = ctx->priv; const int depth = 12; - DO_COMMON(uint16_t, av_clip_uintp2, 1, 1) + DO_COMMON(uint16_t, int, av_clip_uintp2, 1, 1) return 0; } @@ -290,14 +300,21 @@ static int colorlevels_preserve_slice_14_planar(AVFilterContext *ctx, void *arg, { ColorLevelsContext *s = ctx->priv; const int depth = 14; - DO_COMMON(uint16_t, av_clip_uintp2, 1, 1) + DO_COMMON(uint16_t, int, av_clip_uintp2, 1, 1) return 0; } static int colorlevels_preserve_slice_16_planar(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) { ColorLevelsContext *s = ctx->priv; - DO_COMMON(uint16_t, CLIP16, 1, 1) + DO_COMMON(uint16_t, int, CLIP16, 1, 1) + return 0; +} + +static int colorlevels_preserve_slice_32_planar(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) +{ + ColorLevelsContext *s = ctx->priv; + DO_COMMON(float, float, NOCLIP, 1, 1) return 0; } @@ -349,6 +366,10 @@ static int config_input(AVFilterLink *inlink) s->colorlevels_slice[0] = colorlevels_slice_16_planar; s->colorlevels_slice[1] = colorlevels_preserve_slice_16_planar; break; + case 32: + s->colorlevels_slice[0] = colorlevels_slice_32_planar; + s->colorlevels_slice[1] = colorlevels_preserve_slice_32_planar; + break; } } @@ -479,6 +500,46 @@ static int filter_frame(AVFilterLink *inlink, AVFrame *in) td.omin[i] = omin; } break; + case 4: + for (int i = 0; i < s->nb_comp; i++) { + Range *r = &s->range[i]; + const uint8_t offset = s->rgba_map[i]; + const uint8_t *srcrow = in->data[0]; + float imin = r->in_min; + float imax = r->in_max; + float omin = r->out_min; + float omax = r->out_max; + float coeff; + + if (imin < 0.f) { + imin = 1.f; + for (int y = 0; y < inlink->h; y++) { + const float *src = (const float *)srcrow; + + for (int x = 0; x < s->linesize; x += step) + imin = fminf(imin, src[x + offset]); + srcrow += in->linesize[0]; + } + } + if (imax < 0.f) { + srcrow = in->data[0]; + imax = 0.f; + for (int y = 0; y < inlink->h; y++) { + const float *src = (const float *)srcrow; + + for (int x = 0; x < s->linesize; x += step) + imax = fmaxf(imax, src[x + offset]); + srcrow += in->linesize[0]; + } + } + + coeff = (omax - omin) / (double)(imax - imin); + + td.coeff[i] = coeff; + td.fimin[i] = imin; + td.fomin[i] = omin; + } + break; } ff_filter_execute(ctx, s->colorlevels_slice[s->preserve_color > 0], &td, NULL, @@ -524,7 +585,8 @@ const AVFilter ff_vf_colorlevels = { AV_PIX_FMT_GBRP10, AV_PIX_FMT_GBRAP10, AV_PIX_FMT_GBRP12, AV_PIX_FMT_GBRAP12, AV_PIX_FMT_GBRP14, - AV_PIX_FMT_GBRP16, AV_PIX_FMT_GBRAP16), + AV_PIX_FMT_GBRP16, AV_PIX_FMT_GBRAP16, + AV_PIX_FMT_GBRPF32, AV_PIX_FMT_GBRAPF32), .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC | AVFILTER_FLAG_SLICE_THREADS, .process_command = ff_filter_process_command, };

1 0

lavfi/vpp_qsv: factor out the code for filter definition
by Haihao Xiang 30 Jan '23

30 Jan '23

ffmpeg | branch: master | Haihao Xiang <haihao.xiang(a)intel.com> | Tue Jan 17 14:20:27 2023 +0800| [2d202985b79630cd5056c4e32f8f77f22bf1067c] | committer: Haihao Xiang lavfi/vpp_qsv: factor out the code for filter definition Signed-off-by: Haihao Xiang <haihao.xiang(a)intel.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=2d202985b79630cd5… --- libavfilter/vf_vpp_qsv.c | 126 +++++++++++++++++++---------------------------- 1 file changed, 51 insertions(+), 75 deletions(-) diff --git a/libavfilter/vf_vpp_qsv.c b/libavfilter/vf_vpp_qsv.c index 34732bb2c9..34e530e244 100644 --- a/libavfilter/vf_vpp_qsv.c +++ b/libavfilter/vf_vpp_qsv.c @@ -615,36 +615,31 @@ static const AVFilterPad vpp_outputs[] = { }, }; -#if CONFIG_VPP_QSV_FILTER - -static int query_formats(AVFilterContext *ctx) -{ - int ret; - static const enum AVPixelFormat in_pix_fmts[] = { - AV_PIX_FMT_YUV420P, - AV_PIX_FMT_NV12, - AV_PIX_FMT_YUYV422, - AV_PIX_FMT_RGB32, - AV_PIX_FMT_P010, - AV_PIX_FMT_QSV, - AV_PIX_FMT_NONE - }; - static const enum AVPixelFormat out_pix_fmts[] = { - AV_PIX_FMT_NV12, - AV_PIX_FMT_P010, - AV_PIX_FMT_QSV, - AV_PIX_FMT_NONE - }; +#define DEFINE_QSV_FILTER(x, sn, ln, fmts) \ +static const AVClass x##_class = { \ + .class_name = #sn "_qsv", \ + .item_name = av_default_item_name, \ + .option = x##_options, \ + .version = LIBAVUTIL_VERSION_INT, \ +}; \ +const AVFilter ff_vf_##sn##_qsv = { \ + .name = #sn "_qsv", \ + .description = NULL_IF_CONFIG_SMALL("Quick Sync Video " #ln), \ + .preinit = x##_preinit, \ + .init = vpp_init, \ + .uninit = vpp_uninit, \ + .priv_size = sizeof(VPPContext), \ + .priv_class = &x##_class, \ + FILTER_INPUTS(vpp_inputs), \ + FILTER_OUTPUTS(vpp_outputs), \ + fmts, \ + .activate = activate, \ + .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE, \ +}; - ret = ff_formats_ref(ff_make_format_list(in_pix_fmts), - &ctx->inputs[0]->outcfg.formats); - if (ret < 0) - return ret; - return ff_formats_ref(ff_make_format_list(out_pix_fmts), - &ctx->outputs[0]->incfg.formats); -} +#if CONFIG_VPP_QSV_FILTER -static const AVOption options[] = { +static const AVOption vpp_options[] = { { "deinterlace", "deinterlace mode: 0=off, 1=bob, 2=advanced", OFFSET(deinterlace), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, MFX_DEINTERLACING_ADVANCED, .flags = FLAGS, "deinterlace" }, { "bob", "Bob deinterlace mode.", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_DEINTERLACING_BOB }, .flags = FLAGS, "deinterlace" }, { "advanced", "Advanced deinterlace mode. ", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_DEINTERLACING_ADVANCED }, .flags = FLAGS, "deinterlace" }, @@ -686,27 +681,34 @@ static const AVOption options[] = { { NULL } }; -static const AVClass vpp_class = { - .class_name = "vpp_qsv", - .item_name = av_default_item_name, - .option = options, - .version = LIBAVUTIL_VERSION_INT, -}; +static int vpp_query_formats(AVFilterContext *ctx) +{ + int ret; + static const enum AVPixelFormat in_pix_fmts[] = { + AV_PIX_FMT_YUV420P, + AV_PIX_FMT_NV12, + AV_PIX_FMT_YUYV422, + AV_PIX_FMT_RGB32, + AV_PIX_FMT_P010, + AV_PIX_FMT_QSV, + AV_PIX_FMT_NONE + }; + static const enum AVPixelFormat out_pix_fmts[] = { + AV_PIX_FMT_NV12, + AV_PIX_FMT_P010, + AV_PIX_FMT_QSV, + AV_PIX_FMT_NONE + }; -const AVFilter ff_vf_vpp_qsv = { - .name = "vpp_qsv", - .description = NULL_IF_CONFIG_SMALL("Quick Sync Video VPP."), - .priv_size = sizeof(VPPContext), - .preinit = vpp_preinit, - .init = vpp_init, - .uninit = vpp_uninit, - FILTER_INPUTS(vpp_inputs), - FILTER_OUTPUTS(vpp_outputs), - FILTER_QUERY_FUNC(query_formats), - .activate = activate, - .priv_class = &vpp_class, - .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE, -}; + ret = ff_formats_ref(ff_make_format_list(in_pix_fmts), + &ctx->inputs[0]->outcfg.formats); + if (ret < 0) + return ret; + return ff_formats_ref(ff_make_format_list(out_pix_fmts), + &ctx->outputs[0]->incfg.formats); +} + +DEFINE_QSV_FILTER(vpp, vpp, "VPP", FILTER_QUERY_FUNC(vpp_query_formats)); #endif @@ -734,32 +736,6 @@ static av_cold int qsvscale_preinit(AVFilterContext *ctx) return 0; } -static const AVClass qsvscale_class = { - .class_name = "scale_qsv", - .item_name = av_default_item_name, - .option = qsvscale_options, - .version = LIBAVUTIL_VERSION_INT, -}; - -const AVFilter ff_vf_scale_qsv = { - .name = "scale_qsv", - .description = NULL_IF_CONFIG_SMALL("Quick Sync Video scaling and format conversion"), - - .preinit = qsvscale_preinit, - .init = vpp_init, - .uninit = vpp_uninit, - - .priv_size = sizeof(VPPContext), - .priv_class = &qsvscale_class, - - FILTER_INPUTS(vpp_inputs), - FILTER_OUTPUTS(vpp_outputs), - - FILTER_SINGLE_PIXFMT(AV_PIX_FMT_QSV), - - .activate = activate, - - .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE, -}; +DEFINE_QSV_FILTER(qsvscale, scale, "scaling and format conversion", FILTER_SINGLE_PIXFMT(AV_PIX_FMT_QSV)); #endif

1 0

lavfi/scale_qsv: re-use VPPContext for scale_qsv filter
by Haihao Xiang 30 Jan '23

30 Jan '23

ffmpeg | branch: master | Haihao Xiang <haihao.xiang(a)intel.com> | Tue Jan 17 14:20:26 2023 +0800| [9c09fb232b6863181ac1fe70647f9cbccb053d88] | committer: Haihao Xiang lavfi/scale_qsv: re-use VPPContext for scale_qsv filter QSVScaleContext and VPPContext have the same base context, and all features in scale_qsv are implemented in vpp_qsv filter, so scale_qsv can be taken as a special case of vpp_qsv filter, and we may use VPPContext with a different option array, preinit callback and supported pixel formats to implement scale_qsv then remove QSVScaleContext Signed-off-by: Haihao Xiang <haihao.xiang(a)intel.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=9c09fb232b6863181… --- libavfilter/Makefile | 2 +- libavfilter/vf_scale_qsv.c | 290 --------------------------------------------- libavfilter/vf_vpp_qsv.c | 186 +++++++++++++++++++---------- 3 files changed, 124 insertions(+), 354 deletions(-) diff --git a/libavfilter/Makefile b/libavfilter/Makefile index a0059c69e6..b45dcd00fc 100644 --- a/libavfilter/Makefile +++ b/libavfilter/Makefile @@ -446,7 +446,7 @@ OBJS-$(CONFIG_SCALE_FILTER) += vf_scale.o scale_eval.o OBJS-$(CONFIG_SCALE_CUDA_FILTER) += vf_scale_cuda.o scale_eval.o \ vf_scale_cuda.ptx.o cuda/load_helper.o OBJS-$(CONFIG_SCALE_NPP_FILTER) += vf_scale_npp.o scale_eval.o -OBJS-$(CONFIG_SCALE_QSV_FILTER) += vf_scale_qsv.o +OBJS-$(CONFIG_SCALE_QSV_FILTER) += vf_vpp_qsv.o OBJS-$(CONFIG_SCALE_VAAPI_FILTER) += vf_scale_vaapi.o scale_eval.o vaapi_vpp.o OBJS-$(CONFIG_SCALE_VULKAN_FILTER) += vf_scale_vulkan.o vulkan.o vulkan_filter.o OBJS-$(CONFIG_SCALE2REF_FILTER) += vf_scale.o scale_eval.o diff --git a/libavfilter/vf_scale_qsv.c b/libavfilter/vf_scale_qsv.c deleted file mode 100644 index 8eb8bc9ec1..0000000000 --- a/libavfilter/vf_scale_qsv.c +++ /dev/null @@ -1,290 +0,0 @@ -/* - * 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 - */ - -/** - * @file - * scale video filter - QSV - */ - -#include <mfxvideo.h> - -#include <stdio.h> -#include <string.h> - -#include "libavutil/avstring.h" -#include "libavutil/common.h" -#include "libavutil/eval.h" -#include "libavutil/hwcontext.h" -#include "libavutil/hwcontext_qsv.h" -#include "libavutil/internal.h" -#include "libavutil/mathematics.h" -#include "libavutil/opt.h" -#include "libavutil/pixdesc.h" -#include "libavutil/time.h" -#include "libavfilter/qsvvpp.h" - -#include "avfilter.h" -#include "formats.h" -#include "internal.h" -#include "video.h" - -static const char *const var_names[] = { - "in_w", "iw", - "in_h", "ih", - "out_w", "ow", - "out_h", "oh", - "a", "dar", - "sar", - NULL -}; - -enum var_name { - VAR_IN_W, VAR_IW, - VAR_IN_H, VAR_IH, - VAR_OUT_W, VAR_OW, - VAR_OUT_H, VAR_OH, - VAR_A, VAR_DAR, - VAR_SAR, - VARS_NB -}; - -#define MFX_IMPL_VIA_MASK(impl) (0x0f00 & (impl)) - -typedef struct QSVScaleContext { - QSVVPPContext qsv; - - mfxExtVPPScaling scale_conf; - int mode; - - /** - * New dimensions. Special values are: - * 0 = original width/height - * -1 = keep original aspect - */ - int w, h; - - /** - * Output sw format. AV_PIX_FMT_NONE for no conversion. - */ - enum AVPixelFormat format; - - char *w_expr; ///< width expression string - char *h_expr; ///< height expression string - char *format_str; -} QSVScaleContext; - -static av_cold int qsvscale_init(AVFilterContext *ctx) -{ - QSVScaleContext *s = ctx->priv; - - if (!strcmp(s->format_str, "same")) { - s->format = AV_PIX_FMT_NONE; - } else { - s->format = av_get_pix_fmt(s->format_str); - if (s->format == AV_PIX_FMT_NONE) { - av_log(ctx, AV_LOG_ERROR, "Unrecognized pixel format: %s\n", s->format_str); - return AVERROR(EINVAL); - } - } - - return 0; -} - -static av_cold void qsvscale_uninit(AVFilterContext *ctx) -{ - ff_qsvvpp_close(ctx); -} - -static int qsvscale_config_props(AVFilterLink *outlink) -{ - AVFilterContext *ctx = outlink->src; - AVFilterLink *inlink = outlink->src->inputs[0]; - QSVScaleContext *s = ctx->priv; - QSVVPPParam param = { NULL }; - mfxExtBuffer *ext_buf[1]; - int64_t w, h; - double var_values[VARS_NB], res; - char *expr; - int ret; - enum AVPixelFormat in_format; - - var_values[VAR_IN_W] = var_values[VAR_IW] = inlink->w; - var_values[VAR_IN_H] = var_values[VAR_IH] = inlink->h; - var_values[VAR_OUT_W] = var_values[VAR_OW] = NAN; - var_values[VAR_OUT_H] = var_values[VAR_OH] = NAN; - var_values[VAR_A] = (double) inlink->w / inlink->h; - var_values[VAR_SAR] = inlink->sample_aspect_ratio.num ? - (double) inlink->sample_aspect_ratio.num / inlink->sample_aspect_ratio.den : 1; - var_values[VAR_DAR] = var_values[VAR_A] * var_values[VAR_SAR]; - - /* evaluate width and height */ - av_expr_parse_and_eval(&res, (expr = s->w_expr), - var_names, var_values, - NULL, NULL, NULL, NULL, NULL, 0, ctx); - s->w = var_values[VAR_OUT_W] = var_values[VAR_OW] = res; - if ((ret = av_expr_parse_and_eval(&res, (expr = s->h_expr), - var_names, var_values, - NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0) - goto fail; - s->h = var_values[VAR_OUT_H] = var_values[VAR_OH] = res; - /* evaluate again the width, as it may depend on the output height */ - if ((ret = av_expr_parse_and_eval(&res, (expr = s->w_expr), - var_names, var_values, - NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0) - goto fail; - s->w = res; - - w = s->w; - h = s->h; - - /* sanity check params */ - if (w < -1 || h < -1) { - av_log(ctx, AV_LOG_ERROR, "Size values less than -1 are not acceptable.\n"); - return AVERROR(EINVAL); - } - if (w == -1 && h == -1) - s->w = s->h = 0; - - if (!(w = s->w)) - w = inlink->w; - if (!(h = s->h)) - h = inlink->h; - if (w == -1) - w = av_rescale(h, inlink->w, inlink->h); - if (h == -1) - h = av_rescale(w, inlink->h, inlink->w); - - if (w > INT_MAX || h > INT_MAX || - (h * inlink->w) > INT_MAX || - (w * inlink->h) > INT_MAX) - av_log(ctx, AV_LOG_ERROR, "Rescaled value for width or height is too big.\n"); - - outlink->w = w; - outlink->h = h; - - if (inlink->format == AV_PIX_FMT_QSV) { - if (!inlink->hw_frames_ctx || !inlink->hw_frames_ctx->data) - return AVERROR(EINVAL); - else - in_format = ((AVHWFramesContext*)inlink->hw_frames_ctx->data)->sw_format; - } else - in_format = inlink->format; - - if (s->format == AV_PIX_FMT_NONE) - s->format = in_format; - - outlink->frame_rate = inlink->frame_rate; - outlink->time_base = av_inv_q(inlink->frame_rate); - param.out_sw_format = s->format; - - param.ext_buf = ext_buf; - memset(&s->scale_conf, 0, sizeof(mfxExtVPPScaling)); - s->scale_conf.Header.BufferId = MFX_EXTBUFF_VPP_SCALING; - s->scale_conf.Header.BufferSz = sizeof(mfxExtVPPScaling); - s->scale_conf.ScalingMode = s->mode; - param.ext_buf[param.num_ext_buf++] = (mfxExtBuffer*)&s->scale_conf; - av_log(ctx, AV_LOG_VERBOSE, "Scaling mode: %d\n", s->mode); - - ret = ff_qsvvpp_init(ctx, &param); - if (ret < 0) - return ret; - - av_log(ctx, AV_LOG_VERBOSE, "w:%d h:%d -> w:%d h:%d\n", - inlink->w, inlink->h, outlink->w, outlink->h); - - if (inlink->sample_aspect_ratio.num) - outlink->sample_aspect_ratio = av_mul_q((AVRational){outlink->h*inlink->w, - outlink->w*inlink->h}, - inlink->sample_aspect_ratio); - else - outlink->sample_aspect_ratio = inlink->sample_aspect_ratio; - - return 0; - -fail: - av_log(ctx, AV_LOG_ERROR, - "Error when evaluating the expression '%s'\n", expr); - return ret; -} - -static int qsvscale_filter_frame(AVFilterLink *link, AVFrame *in) -{ - int ret = 0; - AVFilterContext *ctx = link->dst; - QSVVPPContext *qsv = ctx->priv; - - ret = ff_qsvvpp_filter_frame(qsv, link, in); - av_frame_free(&in); - return ret; -} - -#define OFFSET(x) offsetof(QSVScaleContext, x) -#define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM -static const AVOption options[] = { - { "w", "Output video width", OFFSET(w_expr), AV_OPT_TYPE_STRING, { .str = "iw" }, .flags = FLAGS }, - { "h", "Output video height", OFFSET(h_expr), AV_OPT_TYPE_STRING, { .str = "ih" }, .flags = FLAGS }, - { "format", "Output pixel format", OFFSET(format_str), AV_OPT_TYPE_STRING, { .str = "same" }, .flags = FLAGS }, - - { "mode", "set scaling mode", OFFSET(mode), AV_OPT_TYPE_INT, { .i64 = MFX_SCALING_MODE_DEFAULT}, MFX_SCALING_MODE_DEFAULT, MFX_SCALING_MODE_QUALITY, FLAGS, "mode"}, - { "low_power", "low power mode", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_SCALING_MODE_LOWPOWER}, INT_MIN, INT_MAX, FLAGS, "mode"}, - { "hq", "high quality mode", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_SCALING_MODE_QUALITY}, INT_MIN, INT_MAX, FLAGS, "mode"}, - - { NULL }, -}; - -static const AVClass qsvscale_class = { - .class_name = "scale_qsv", - .item_name = av_default_item_name, - .option = options, - .version = LIBAVUTIL_VERSION_INT, -}; - -static const AVFilterPad qsvscale_inputs[] = { - { - .name = "default", - .type = AVMEDIA_TYPE_VIDEO, - .filter_frame = qsvscale_filter_frame, - .get_buffer.video = ff_qsvvpp_get_video_buffer, - }, -}; - -static const AVFilterPad qsvscale_outputs[] = { - { - .name = "default", - .type = AVMEDIA_TYPE_VIDEO, - .config_props = qsvscale_config_props, - }, -}; - -const AVFilter ff_vf_scale_qsv = { - .name = "scale_qsv", - .description = NULL_IF_CONFIG_SMALL("QuickSync video scaling and format conversion"), - - .init = qsvscale_init, - .uninit = qsvscale_uninit, - - .priv_size = sizeof(QSVScaleContext), - .priv_class = &qsvscale_class, - - FILTER_INPUTS(qsvscale_inputs), - FILTER_OUTPUTS(qsvscale_outputs), - - FILTER_SINGLE_PIXFMT(AV_PIX_FMT_QSV), - - .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE, -}; diff --git a/libavfilter/vf_vpp_qsv.c b/libavfilter/vf_vpp_qsv.c index 4a1af7146a..34732bb2c9 100644 --- a/libavfilter/vf_vpp_qsv.c +++ b/libavfilter/vf_vpp_qsv.c @@ -23,6 +23,8 @@ #include <float.h> +#include "config_components.h" + #include "libavutil/opt.h" #include "libavutil/eval.h" #include "libavutil/hwcontext.h" @@ -101,48 +103,6 @@ typedef struct VPPContext{ int has_passthrough; /* apply pass through mode if possible */ } VPPContext; -static const AVOption options[] = { - { "deinterlace", "deinterlace mode: 0=off, 1=bob, 2=advanced", OFFSET(deinterlace), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, MFX_DEINTERLACING_ADVANCED, .flags = FLAGS, "deinterlace" }, - { "bob", "Bob deinterlace mode.", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_DEINTERLACING_BOB }, .flags = FLAGS, "deinterlace" }, - { "advanced", "Advanced deinterlace mode. ", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_DEINTERLACING_ADVANCED }, .flags = FLAGS, "deinterlace" }, - - { "denoise", "denoise level [0, 100]", OFFSET(denoise), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 100, .flags = FLAGS }, - { "detail", "enhancement level [0, 100]", OFFSET(detail), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 100, .flags = FLAGS }, - { "framerate", "output framerate", OFFSET(framerate), AV_OPT_TYPE_RATIONAL, { .dbl = 0.0 },0, DBL_MAX, .flags = FLAGS }, - { "procamp", "Enable ProcAmp", OFFSET(procamp), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, .flags = FLAGS}, - { "hue", "ProcAmp hue", OFFSET(hue), AV_OPT_TYPE_FLOAT, { .dbl = 0.0 }, -180.0, 180.0, .flags = FLAGS}, - { "saturation", "ProcAmp saturation", OFFSET(saturation), AV_OPT_TYPE_FLOAT, { .dbl = 1.0 }, 0.0, 10.0, .flags = FLAGS}, - { "contrast", "ProcAmp contrast", OFFSET(contrast), AV_OPT_TYPE_FLOAT, { .dbl = 1.0 }, 0.0, 10.0, .flags = FLAGS}, - { "brightness", "ProcAmp brightness", OFFSET(brightness), AV_OPT_TYPE_FLOAT, { .dbl = 0.0 }, -100.0, 100.0, .flags = FLAGS}, - - { "transpose", "set transpose direction", OFFSET(transpose), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 6, FLAGS, "transpose"}, - { "cclock_hflip", "rotate counter-clockwise with horizontal flip", 0, AV_OPT_TYPE_CONST, { .i64 = TRANSPOSE_CCLOCK_FLIP }, .flags=FLAGS, .unit = "transpose" }, - { "clock", "rotate clockwise", 0, AV_OPT_TYPE_CONST, { .i64 = TRANSPOSE_CLOCK }, .flags=FLAGS, .unit = "transpose" }, - { "cclock", "rotate counter-clockwise", 0, AV_OPT_TYPE_CONST, { .i64 = TRANSPOSE_CCLOCK }, .flags=FLAGS, .unit = "transpose" }, - { "clock_hflip", "rotate clockwise with horizontal flip", 0, AV_OPT_TYPE_CONST, { .i64 = TRANSPOSE_CLOCK_FLIP }, .flags=FLAGS, .unit = "transpose" }, - { "reversal", "rotate by half-turn", 0, AV_OPT_TYPE_CONST, { .i64 = TRANSPOSE_REVERSAL }, .flags=FLAGS, .unit = "transpose" }, - { "hflip", "flip horizontally", 0, AV_OPT_TYPE_CONST, { .i64 = TRANSPOSE_HFLIP }, .flags=FLAGS, .unit = "transpose" }, - { "vflip", "flip vertically", 0, AV_OPT_TYPE_CONST, { .i64 = TRANSPOSE_VFLIP }, .flags=FLAGS, .unit = "transpose" }, - - { "cw", "set the width crop area expression", OFFSET(cw), AV_OPT_TYPE_STRING, { .str = "iw" }, 0, 0, FLAGS }, - { "ch", "set the height crop area expression", OFFSET(ch), AV_OPT_TYPE_STRING, { .str = "ih" }, 0, 0, FLAGS }, - { "cx", "set the x crop area expression", OFFSET(cx), AV_OPT_TYPE_STRING, { .str = "(in_w-out_w)/2" }, 0, 0, FLAGS }, - { "cy", "set the y crop area expression", OFFSET(cy), AV_OPT_TYPE_STRING, { .str = "(in_h-out_h)/2" }, 0, 0, FLAGS }, - - { "w", "Output video width(0=input video width, -1=keep input video aspect)", OFFSET(ow), AV_OPT_TYPE_STRING, { .str="cw" }, 0, 255, .flags = FLAGS }, - { "width", "Output video width(0=input video width, -1=keep input video aspect)", OFFSET(ow), AV_OPT_TYPE_STRING, { .str="cw" }, 0, 255, .flags = FLAGS }, - { "h", "Output video height(0=input video height, -1=keep input video aspect)", OFFSET(oh), AV_OPT_TYPE_STRING, { .str="w*ch/cw" }, 0, 255, .flags = FLAGS }, - { "height", "Output video height(0=input video height, -1=keep input video aspect)", OFFSET(oh), AV_OPT_TYPE_STRING, { .str="w*ch/cw" }, 0, 255, .flags = FLAGS }, - { "format", "Output pixel format", OFFSET(output_format_str), AV_OPT_TYPE_STRING, { .str = "same" }, .flags = FLAGS }, - { "async_depth", "Internal parallelization depth, the higher the value the higher the latency.", OFFSET(qsv.async_depth), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, .flags = FLAGS }, - { "scale_mode", "scale & format conversion mode: 0=auto, 1=low power, 2=high quality", OFFSET(scale_mode), AV_OPT_TYPE_INT, { .i64 = MFX_SCALING_MODE_DEFAULT }, MFX_SCALING_MODE_DEFAULT, MFX_SCALING_MODE_QUALITY, .flags = FLAGS, "scale mode" }, - { "auto", "auto mode", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_SCALING_MODE_DEFAULT}, INT_MIN, INT_MAX, FLAGS, "scale mode"}, - { "low_power", "low power mode", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_SCALING_MODE_LOWPOWER}, INT_MIN, INT_MAX, FLAGS, "scale mode"}, - { "hq", "high quality mode", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_SCALING_MODE_QUALITY}, INT_MIN, INT_MAX, FLAGS, "scale mode"}, - - { NULL } -}; - static const char *const var_names[] = { "iw", "in_w", "ih", "in_h", @@ -633,6 +593,30 @@ eof: return 0; } +static av_cold void vpp_uninit(AVFilterContext *ctx) +{ + ff_qsvvpp_close(ctx); +} + +static const AVFilterPad vpp_inputs[] = { + { + .name = "default", + .type = AVMEDIA_TYPE_VIDEO, + .config_props = config_input, + .get_buffer.video = ff_qsvvpp_get_video_buffer, + }, +}; + +static const AVFilterPad vpp_outputs[] = { + { + .name = "default", + .type = AVMEDIA_TYPE_VIDEO, + .config_props = config_output, + }, +}; + +#if CONFIG_VPP_QSV_FILTER + static int query_formats(AVFilterContext *ctx) { int ret; @@ -660,10 +644,47 @@ static int query_formats(AVFilterContext *ctx) &ctx->outputs[0]->incfg.formats); } -static av_cold void vpp_uninit(AVFilterContext *ctx) -{ - ff_qsvvpp_close(ctx); -} +static const AVOption options[] = { + { "deinterlace", "deinterlace mode: 0=off, 1=bob, 2=advanced", OFFSET(deinterlace), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, MFX_DEINTERLACING_ADVANCED, .flags = FLAGS, "deinterlace" }, + { "bob", "Bob deinterlace mode.", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_DEINTERLACING_BOB }, .flags = FLAGS, "deinterlace" }, + { "advanced", "Advanced deinterlace mode. ", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_DEINTERLACING_ADVANCED }, .flags = FLAGS, "deinterlace" }, + + { "denoise", "denoise level [0, 100]", OFFSET(denoise), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 100, .flags = FLAGS }, + { "detail", "enhancement level [0, 100]", OFFSET(detail), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 100, .flags = FLAGS }, + { "framerate", "output framerate", OFFSET(framerate), AV_OPT_TYPE_RATIONAL, { .dbl = 0.0 },0, DBL_MAX, .flags = FLAGS }, + { "procamp", "Enable ProcAmp", OFFSET(procamp), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, .flags = FLAGS}, + { "hue", "ProcAmp hue", OFFSET(hue), AV_OPT_TYPE_FLOAT, { .dbl = 0.0 }, -180.0, 180.0, .flags = FLAGS}, + { "saturation", "ProcAmp saturation", OFFSET(saturation), AV_OPT_TYPE_FLOAT, { .dbl = 1.0 }, 0.0, 10.0, .flags = FLAGS}, + { "contrast", "ProcAmp contrast", OFFSET(contrast), AV_OPT_TYPE_FLOAT, { .dbl = 1.0 }, 0.0, 10.0, .flags = FLAGS}, + { "brightness", "ProcAmp brightness", OFFSET(brightness), AV_OPT_TYPE_FLOAT, { .dbl = 0.0 }, -100.0, 100.0, .flags = FLAGS}, + + { "transpose", "set transpose direction", OFFSET(transpose), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 6, FLAGS, "transpose"}, + { "cclock_hflip", "rotate counter-clockwise with horizontal flip", 0, AV_OPT_TYPE_CONST, { .i64 = TRANSPOSE_CCLOCK_FLIP }, .flags=FLAGS, .unit = "transpose" }, + { "clock", "rotate clockwise", 0, AV_OPT_TYPE_CONST, { .i64 = TRANSPOSE_CLOCK }, .flags=FLAGS, .unit = "transpose" }, + { "cclock", "rotate counter-clockwise", 0, AV_OPT_TYPE_CONST, { .i64 = TRANSPOSE_CCLOCK }, .flags=FLAGS, .unit = "transpose" }, + { "clock_hflip", "rotate clockwise with horizontal flip", 0, AV_OPT_TYPE_CONST, { .i64 = TRANSPOSE_CLOCK_FLIP }, .flags=FLAGS, .unit = "transpose" }, + { "reversal", "rotate by half-turn", 0, AV_OPT_TYPE_CONST, { .i64 = TRANSPOSE_REVERSAL }, .flags=FLAGS, .unit = "transpose" }, + { "hflip", "flip horizontally", 0, AV_OPT_TYPE_CONST, { .i64 = TRANSPOSE_HFLIP }, .flags=FLAGS, .unit = "transpose" }, + { "vflip", "flip vertically", 0, AV_OPT_TYPE_CONST, { .i64 = TRANSPOSE_VFLIP }, .flags=FLAGS, .unit = "transpose" }, + + { "cw", "set the width crop area expression", OFFSET(cw), AV_OPT_TYPE_STRING, { .str = "iw" }, 0, 0, FLAGS }, + { "ch", "set the height crop area expression", OFFSET(ch), AV_OPT_TYPE_STRING, { .str = "ih" }, 0, 0, FLAGS }, + { "cx", "set the x crop area expression", OFFSET(cx), AV_OPT_TYPE_STRING, { .str = "(in_w-out_w)/2" }, 0, 0, FLAGS }, + { "cy", "set the y crop area expression", OFFSET(cy), AV_OPT_TYPE_STRING, { .str = "(in_h-out_h)/2" }, 0, 0, FLAGS }, + + { "w", "Output video width(0=input video width, -1=keep input video aspect)", OFFSET(ow), AV_OPT_TYPE_STRING, { .str="cw" }, 0, 255, .flags = FLAGS }, + { "width", "Output video width(0=input video width, -1=keep input video aspect)", OFFSET(ow), AV_OPT_TYPE_STRING, { .str="cw" }, 0, 255, .flags = FLAGS }, + { "h", "Output video height(0=input video height, -1=keep input video aspect)", OFFSET(oh), AV_OPT_TYPE_STRING, { .str="w*ch/cw" }, 0, 255, .flags = FLAGS }, + { "height", "Output video height(0=input video height, -1=keep input video aspect)", OFFSET(oh), AV_OPT_TYPE_STRING, { .str="w*ch/cw" }, 0, 255, .flags = FLAGS }, + { "format", "Output pixel format", OFFSET(output_format_str), AV_OPT_TYPE_STRING, { .str = "same" }, .flags = FLAGS }, + { "async_depth", "Internal parallelization depth, the higher the value the higher the latency.", OFFSET(qsv.async_depth), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, .flags = FLAGS }, + { "scale_mode", "scale & format conversion mode: 0=auto, 1=low power, 2=high quality", OFFSET(scale_mode), AV_OPT_TYPE_INT, { .i64 = MFX_SCALING_MODE_DEFAULT }, MFX_SCALING_MODE_DEFAULT, MFX_SCALING_MODE_QUALITY, .flags = FLAGS, "scale mode" }, + { "auto", "auto mode", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_SCALING_MODE_DEFAULT}, INT_MIN, INT_MAX, FLAGS, "scale mode"}, + { "low_power", "low power mode", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_SCALING_MODE_LOWPOWER}, INT_MIN, INT_MAX, FLAGS, "scale mode"}, + { "hq", "high quality mode", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_SCALING_MODE_QUALITY}, INT_MIN, INT_MAX, FLAGS, "scale mode"}, + + { NULL } +}; static const AVClass vpp_class = { .class_name = "vpp_qsv", @@ -672,23 +693,6 @@ static const AVClass vpp_class = { .version = LIBAVUTIL_VERSION_INT, }; -static const AVFilterPad vpp_inputs[] = { - { - .name = "default", - .type = AVMEDIA_TYPE_VIDEO, - .config_props = config_input, - .get_buffer.video = ff_qsvvpp_get_video_buffer, - }, -}; - -static const AVFilterPad vpp_outputs[] = { - { - .name = "default", - .type = AVMEDIA_TYPE_VIDEO, - .config_props = config_output, - }, -}; - const AVFilter ff_vf_vpp_qsv = { .name = "vpp_qsv", .description = NULL_IF_CONFIG_SMALL("Quick Sync Video VPP."), @@ -703,3 +707,59 @@ const AVFilter ff_vf_vpp_qsv = { .priv_class = &vpp_class, .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE, }; + +#endif + +#if CONFIG_SCALE_QSV_FILTER + +static const AVOption qsvscale_options[] = { + { "w", "Output video width(0=input video width, -1=keep input video aspect)", OFFSET(ow), AV_OPT_TYPE_STRING, { .str = "iw" }, .flags = FLAGS }, + { "h", "Output video height(0=input video height, -1=keep input video aspect)", OFFSET(oh), AV_OPT_TYPE_STRING, { .str = "ih" }, .flags = FLAGS }, + { "format", "Output pixel format", OFFSET(output_format_str), AV_OPT_TYPE_STRING, { .str = "same" }, .flags = FLAGS }, + + { "mode", "set scaling mode", OFFSET(scale_mode), AV_OPT_TYPE_INT, { .i64 = MFX_SCALING_MODE_DEFAULT}, MFX_SCALING_MODE_DEFAULT, MFX_SCALING_MODE_QUALITY, FLAGS, "mode"}, + { "low_power", "low power mode", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_SCALING_MODE_LOWPOWER}, INT_MIN, INT_MAX, FLAGS, "mode"}, + { "hq", "high quality mode", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_SCALING_MODE_QUALITY}, INT_MIN, INT_MAX, FLAGS, "mode"}, + + { NULL }, +}; + +static av_cold int qsvscale_preinit(AVFilterContext *ctx) +{ + VPPContext *vpp = ctx->priv; + + vpp_preinit(ctx); + vpp->has_passthrough = 0; + + return 0; +} + +static const AVClass qsvscale_class = { + .class_name = "scale_qsv", + .item_name = av_default_item_name, + .option = qsvscale_options, + .version = LIBAVUTIL_VERSION_INT, +}; + +const AVFilter ff_vf_scale_qsv = { + .name = "scale_qsv", + .description = NULL_IF_CONFIG_SMALL("Quick Sync Video scaling and format conversion"), + + .preinit = qsvscale_preinit, + .init = vpp_init, + .uninit = vpp_uninit, + + .priv_size = sizeof(VPPContext), + .priv_class = &qsvscale_class, + + FILTER_INPUTS(vpp_inputs), + FILTER_OUTPUTS(vpp_outputs), + + FILTER_SINGLE_PIXFMT(AV_PIX_FMT_QSV), + + .activate = activate, + + .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE, +}; + +#endif

1 0

lavfi/scale_qsv: simplify scale_qsv filter
by Haihao Xiang 30 Jan '23

30 Jan '23

ffmpeg | branch: master | Haihao Xiang <haihao.xiang(a)intel.com> | Tue Jan 17 14:20:25 2023 +0800| [d66312cf278f0b2db4d4410dd6a99bb4d1f30919] | committer: Haihao Xiang lavfi/scale_qsv: simplify scale_qsv filter Use QSVVPPContext as a base context of QSVScaleContext, hence we may re-use functions defined for QSVVPPContext to manage MFX session for scale_qsv filter. In addition, system memory has been taken into account in QSVVVPPContext, we may add support for non-QSV pixel formats in the future. Signed-off-by: Haihao Xiang <haihao.xiang(a)intel.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=d66312cf278f0b2db… --- configure | 1 + libavfilter/vf_scale_qsv.c | 441 ++++----------------------------------------- 2 files changed, 35 insertions(+), 407 deletions(-) diff --git a/configure b/configure index 8190e978a8..47790d10f5 100755 --- a/configure +++ b/configure @@ -3722,6 +3722,7 @@ sab_filter_deps="gpl swscale" scale2ref_filter_deps="swscale" scale_filter_deps="swscale" scale_qsv_filter_deps="libmfx" +scale_qsv_filter_select="qsvvpp" scdet_filter_select="scene_sad" select_filter_select="scene_sad" sharpness_vaapi_filter_deps="vaapi" diff --git a/libavfilter/vf_scale_qsv.c b/libavfilter/vf_scale_qsv.c index a89a3ba6e6..8eb8bc9ec1 100644 --- a/libavfilter/vf_scale_qsv.c +++ b/libavfilter/vf_scale_qsv.c @@ -66,35 +66,11 @@ enum var_name { #define MFX_IMPL_VIA_MASK(impl) (0x0f00 & (impl)) typedef struct QSVScaleContext { - const AVClass *class; - - /* a clone of the main session, used internally for scaling */ - mfxSession session; - - mfxMemId *mem_ids_in; - int nb_mem_ids_in; - - mfxMemId *mem_ids_out; - int nb_mem_ids_out; - - mfxFrameSurface1 **surface_ptrs_in; - int nb_surface_ptrs_in; - - mfxFrameSurface1 **surface_ptrs_out; - int nb_surface_ptrs_out; - -#if QSV_HAVE_OPAQUE - mfxExtOpaqueSurfaceAlloc opaque_alloc; -#endif + QSVVPPContext qsv; mfxExtVPPScaling scale_conf; int mode; - mfxExtBuffer *ext_buffers[2]; - int num_ext_buf; - - int shift_width, shift_height; - /** * New dimensions. Special values are: * 0 = original width/height @@ -131,338 +107,21 @@ static av_cold int qsvscale_init(AVFilterContext *ctx) static av_cold void qsvscale_uninit(AVFilterContext *ctx) { - QSVScaleContext *s = ctx->priv; - - if (s->session) { - MFXClose(s->session); - s->session = NULL; - } - - av_freep(&s->mem_ids_in); - av_freep(&s->mem_ids_out); - s->nb_mem_ids_in = 0; - s->nb_mem_ids_out = 0; - - av_freep(&s->surface_ptrs_in); - av_freep(&s->surface_ptrs_out); - s->nb_surface_ptrs_in = 0; - s->nb_surface_ptrs_out = 0; -} - -static int init_out_pool(AVFilterContext *ctx, - int out_width, int out_height) -{ - QSVScaleContext *s = ctx->priv; - AVFilterLink *outlink = ctx->outputs[0]; - - AVHWFramesContext *in_frames_ctx; - AVHWFramesContext *out_frames_ctx; - AVQSVFramesContext *in_frames_hwctx; - AVQSVFramesContext *out_frames_hwctx; - enum AVPixelFormat in_format; - enum AVPixelFormat out_format; - int i, ret; - - /* check that we have a hw context */ - if (!ctx->inputs[0]->hw_frames_ctx) { - av_log(ctx, AV_LOG_ERROR, "No hw context provided on input\n"); - return AVERROR(EINVAL); - } - in_frames_ctx = (AVHWFramesContext*)ctx->inputs[0]->hw_frames_ctx->data; - in_frames_hwctx = in_frames_ctx->hwctx; - - in_format = in_frames_ctx->sw_format; - out_format = (s->format == AV_PIX_FMT_NONE) ? in_format : s->format; - - outlink->hw_frames_ctx = av_hwframe_ctx_alloc(in_frames_ctx->device_ref); - if (!outlink->hw_frames_ctx) - return AVERROR(ENOMEM); - out_frames_ctx = (AVHWFramesContext*)outlink->hw_frames_ctx->data; - out_frames_hwctx = out_frames_ctx->hwctx; - - out_frames_ctx->format = AV_PIX_FMT_QSV; - out_frames_ctx->width = FFALIGN(out_width, 16); - out_frames_ctx->height = FFALIGN(out_height, 16); - out_frames_ctx->sw_format = out_format; - out_frames_ctx->initial_pool_size = 4; - - out_frames_hwctx->frame_type = in_frames_hwctx->frame_type | MFX_MEMTYPE_FROM_VPPOUT; - - ret = ff_filter_init_hw_frames(ctx, outlink, 32); - if (ret < 0) - return ret; - - ret = av_hwframe_ctx_init(outlink->hw_frames_ctx); - if (ret < 0) - return ret; - - for (i = 0; i < out_frames_hwctx->nb_surfaces; i++) { - mfxFrameInfo *info = &out_frames_hwctx->surfaces[i].Info; - info->CropW = out_width; - info->CropH = out_height; - } - - return 0; -} - -static mfxStatus frame_alloc(mfxHDL pthis, mfxFrameAllocRequest *req, - mfxFrameAllocResponse *resp) -{ - AVFilterContext *ctx = pthis; - QSVScaleContext *s = ctx->priv; - - if (!(req->Type & MFX_MEMTYPE_VIDEO_MEMORY_PROCESSOR_TARGET) || - !(req->Type & (MFX_MEMTYPE_FROM_VPPIN | MFX_MEMTYPE_FROM_VPPOUT)) || - !(req->Type & MFX_MEMTYPE_EXTERNAL_FRAME)) - return MFX_ERR_UNSUPPORTED; - - if (req->Type & MFX_MEMTYPE_FROM_VPPIN) { - resp->mids = s->mem_ids_in; - resp->NumFrameActual = s->nb_mem_ids_in; - } else { - resp->mids = s->mem_ids_out; - resp->NumFrameActual = s->nb_mem_ids_out; - } - - return MFX_ERR_NONE; -} - -static mfxStatus frame_free(mfxHDL pthis, mfxFrameAllocResponse *resp) -{ - return MFX_ERR_NONE; -} - -static mfxStatus frame_lock(mfxHDL pthis, mfxMemId mid, mfxFrameData *ptr) -{ - return MFX_ERR_UNSUPPORTED; -} - -static mfxStatus frame_unlock(mfxHDL pthis, mfxMemId mid, mfxFrameData *ptr) -{ - return MFX_ERR_UNSUPPORTED; -} - -static mfxStatus frame_get_hdl(mfxHDL pthis, mfxMemId mid, mfxHDL *hdl) -{ - mfxHDLPair *pair_dst = (mfxHDLPair*)hdl; - mfxHDLPair *pair_src = (mfxHDLPair*)mid; - - pair_dst->first = pair_src->first; - - if (pair_src->second != (mfxMemId)MFX_INFINITE) - pair_dst->second = pair_src->second; - return MFX_ERR_NONE; -} - -static int init_out_session(AVFilterContext *ctx) -{ - - QSVScaleContext *s = ctx->priv; - AVHWFramesContext *in_frames_ctx = (AVHWFramesContext*)ctx->inputs[0]->hw_frames_ctx->data; - AVHWFramesContext *out_frames_ctx = (AVHWFramesContext*)ctx->outputs[0]->hw_frames_ctx->data; - AVQSVFramesContext *in_frames_hwctx = in_frames_ctx->hwctx; - AVQSVFramesContext *out_frames_hwctx = out_frames_ctx->hwctx; - AVQSVDeviceContext *device_hwctx = in_frames_ctx->device_ctx->hwctx; - - int opaque = 0; - - mfxHDL handle = NULL; - mfxHandleType handle_type; - mfxVersion ver; - mfxIMPL impl; - mfxVideoParam par; - mfxStatus err; - int i, ret; - -#if QSV_HAVE_OPAQUE - opaque = !!(in_frames_hwctx->frame_type & MFX_MEMTYPE_OPAQUE_FRAME); -#endif - s->num_ext_buf = 0; - - /* extract the properties of the "master" session given to us */ - err = MFXQueryIMPL(device_hwctx->session, &impl); - if (err == MFX_ERR_NONE) - err = MFXQueryVersion(device_hwctx->session, &ver); - if (err != MFX_ERR_NONE) { - av_log(ctx, AV_LOG_ERROR, "Error querying the session attributes\n"); - return AVERROR_UNKNOWN; - } - - if (MFX_IMPL_VIA_VAAPI == MFX_IMPL_VIA_MASK(impl)) { - handle_type = MFX_HANDLE_VA_DISPLAY; - } else if (MFX_IMPL_VIA_D3D11 == MFX_IMPL_VIA_MASK(impl)) { - handle_type = MFX_HANDLE_D3D11_DEVICE; - } else if (MFX_IMPL_VIA_D3D9 == MFX_IMPL_VIA_MASK(impl)) { - handle_type = MFX_HANDLE_D3D9_DEVICE_MANAGER; - } else { - av_log(ctx, AV_LOG_ERROR, "Error unsupported handle type\n"); - return AVERROR_UNKNOWN; - } - - err = MFXVideoCORE_GetHandle(device_hwctx->session, handle_type, &handle); - if (err < 0) - return ff_qsvvpp_print_error(ctx, err, "Error getting the session handle"); - else if (err > 0) { - ff_qsvvpp_print_warning(ctx, err, "Warning in getting the session handle"); - return AVERROR_UNKNOWN; - } - - /* create a "slave" session with those same properties, to be used for - * actual scaling */ - ret = ff_qsvvpp_create_mfx_session(ctx, device_hwctx->loader, impl, &ver, - &s->session); - if (ret) - return ret; - - if (handle) { - err = MFXVideoCORE_SetHandle(s->session, handle_type, handle); - if (err != MFX_ERR_NONE) - return AVERROR_UNKNOWN; - } - - if (QSV_RUNTIME_VERSION_ATLEAST(ver, 1, 25)) { - err = MFXJoinSession(device_hwctx->session, s->session); - if (err != MFX_ERR_NONE) - return AVERROR_UNKNOWN; - } - - memset(&par, 0, sizeof(par)); - - if (!opaque) { - mfxFrameAllocator frame_allocator = { - .pthis = ctx, - .Alloc = frame_alloc, - .Lock = frame_lock, - .Unlock = frame_unlock, - .GetHDL = frame_get_hdl, - .Free = frame_free, - }; - - s->mem_ids_in = av_calloc(in_frames_hwctx->nb_surfaces, - sizeof(*s->mem_ids_in)); - if (!s->mem_ids_in) - return AVERROR(ENOMEM); - for (i = 0; i < in_frames_hwctx->nb_surfaces; i++) - s->mem_ids_in[i] = in_frames_hwctx->surfaces[i].Data.MemId; - s->nb_mem_ids_in = in_frames_hwctx->nb_surfaces; - - s->mem_ids_out = av_calloc(out_frames_hwctx->nb_surfaces, - sizeof(*s->mem_ids_out)); - if (!s->mem_ids_out) - return AVERROR(ENOMEM); - for (i = 0; i < out_frames_hwctx->nb_surfaces; i++) - s->mem_ids_out[i] = out_frames_hwctx->surfaces[i].Data.MemId; - s->nb_mem_ids_out = out_frames_hwctx->nb_surfaces; - - err = MFXVideoCORE_SetFrameAllocator(s->session, &frame_allocator); - if (err != MFX_ERR_NONE) - return AVERROR_UNKNOWN; - - par.IOPattern = MFX_IOPATTERN_IN_VIDEO_MEMORY | MFX_IOPATTERN_OUT_VIDEO_MEMORY; - } -#if QSV_HAVE_OPAQUE - else { - s->surface_ptrs_in = av_calloc(in_frames_hwctx->nb_surfaces, - sizeof(*s->surface_ptrs_in)); - if (!s->surface_ptrs_in) - return AVERROR(ENOMEM); - for (i = 0; i < in_frames_hwctx->nb_surfaces; i++) - s->surface_ptrs_in[i] = in_frames_hwctx->surfaces + i; - s->nb_surface_ptrs_in = in_frames_hwctx->nb_surfaces; - - s->surface_ptrs_out = av_calloc(out_frames_hwctx->nb_surfaces, - sizeof(*s->surface_ptrs_out)); - if (!s->surface_ptrs_out) - return AVERROR(ENOMEM); - for (i = 0; i < out_frames_hwctx->nb_surfaces; i++) - s->surface_ptrs_out[i] = out_frames_hwctx->surfaces + i; - s->nb_surface_ptrs_out = out_frames_hwctx->nb_surfaces; - - s->opaque_alloc.In.Surfaces = s->surface_ptrs_in; - s->opaque_alloc.In.NumSurface = s->nb_surface_ptrs_in; - s->opaque_alloc.In.Type = in_frames_hwctx->frame_type; - - s->opaque_alloc.Out.Surfaces = s->surface_ptrs_out; - s->opaque_alloc.Out.NumSurface = s->nb_surface_ptrs_out; - s->opaque_alloc.Out.Type = out_frames_hwctx->frame_type; - - s->opaque_alloc.Header.BufferId = MFX_EXTBUFF_OPAQUE_SURFACE_ALLOCATION; - s->opaque_alloc.Header.BufferSz = sizeof(s->opaque_alloc); - - s->ext_buffers[s->num_ext_buf++] = (mfxExtBuffer*)&s->opaque_alloc; - - par.IOPattern = MFX_IOPATTERN_IN_OPAQUE_MEMORY | MFX_IOPATTERN_OUT_OPAQUE_MEMORY; - } -#endif - - memset(&s->scale_conf, 0, sizeof(mfxExtVPPScaling)); - s->scale_conf.Header.BufferId = MFX_EXTBUFF_VPP_SCALING; - s->scale_conf.Header.BufferSz = sizeof(mfxExtVPPScaling); - s->scale_conf.ScalingMode = s->mode; - s->ext_buffers[s->num_ext_buf++] = (mfxExtBuffer*)&s->scale_conf; - av_log(ctx, AV_LOG_VERBOSE, "Scaling mode: %d\n", s->mode); - - par.ExtParam = s->ext_buffers; - par.NumExtParam = s->num_ext_buf; - - par.AsyncDepth = 1; // TODO async - - par.vpp.In = in_frames_hwctx->surfaces[0].Info; - par.vpp.Out = out_frames_hwctx->surfaces[0].Info; - - /* Apparently VPP requires the frame rate to be set to some value, otherwise - * init will fail (probably for the framerate conversion filter). Since we - * are only doing scaling here, we just invent an arbitrary - * value */ - par.vpp.In.FrameRateExtN = 25; - par.vpp.In.FrameRateExtD = 1; - par.vpp.Out.FrameRateExtN = 25; - par.vpp.Out.FrameRateExtD = 1; - - /* Print input memory mode */ - ff_qsvvpp_print_iopattern(ctx, par.IOPattern & 0x0F, "VPP"); - /* Print output memory mode */ - ff_qsvvpp_print_iopattern(ctx, par.IOPattern & 0xF0, "VPP"); - err = MFXVideoVPP_Init(s->session, &par); - if (err < 0) - return ff_qsvvpp_print_error(ctx, err, - "Error opening the VPP for scaling"); - else if (err > 0) { - ff_qsvvpp_print_warning(ctx, err, - "Warning in VPP initialization"); - return AVERROR_UNKNOWN; - } - - return 0; -} - -static int init_scale_session(AVFilterContext *ctx, int in_width, int in_height, - int out_width, int out_height) -{ - int ret; - - qsvscale_uninit(ctx); - - ret = init_out_pool(ctx, out_width, out_height); - if (ret < 0) - return ret; - - ret = init_out_session(ctx); - if (ret < 0) - return ret; - - return 0; + ff_qsvvpp_close(ctx); } static int qsvscale_config_props(AVFilterLink *outlink) { AVFilterContext *ctx = outlink->src; AVFilterLink *inlink = outlink->src->inputs[0]; - QSVScaleContext *s = ctx->priv; + QSVScaleContext *s = ctx->priv; + QSVVPPParam param = { NULL }; + mfxExtBuffer *ext_buf[1]; int64_t w, h; double var_values[VARS_NB], res; char *expr; int ret; + enum AVPixelFormat in_format; var_values[VAR_IN_W] = var_values[VAR_IW] = inlink->w; var_values[VAR_IN_H] = var_values[VAR_IH] = inlink->h; @@ -518,7 +177,30 @@ static int qsvscale_config_props(AVFilterLink *outlink) outlink->w = w; outlink->h = h; - ret = init_scale_session(ctx, inlink->w, inlink->h, w, h); + if (inlink->format == AV_PIX_FMT_QSV) { + if (!inlink->hw_frames_ctx || !inlink->hw_frames_ctx->data) + return AVERROR(EINVAL); + else + in_format = ((AVHWFramesContext*)inlink->hw_frames_ctx->data)->sw_format; + } else + in_format = inlink->format; + + if (s->format == AV_PIX_FMT_NONE) + s->format = in_format; + + outlink->frame_rate = inlink->frame_rate; + outlink->time_base = av_inv_q(inlink->frame_rate); + param.out_sw_format = s->format; + + param.ext_buf = ext_buf; + memset(&s->scale_conf, 0, sizeof(mfxExtVPPScaling)); + s->scale_conf.Header.BufferId = MFX_EXTBUFF_VPP_SCALING; + s->scale_conf.Header.BufferSz = sizeof(mfxExtVPPScaling); + s->scale_conf.ScalingMode = s->mode; + param.ext_buf[param.num_ext_buf++] = (mfxExtBuffer*)&s->scale_conf; + av_log(ctx, AV_LOG_VERBOSE, "Scaling mode: %d\n", s->mode); + + ret = ff_qsvvpp_init(ctx, &param); if (ret < 0) return ret; @@ -542,67 +224,12 @@ fail: static int qsvscale_filter_frame(AVFilterLink *link, AVFrame *in) { - AVFilterContext *ctx = link->dst; - QSVScaleContext *s = ctx->priv; - AVFilterLink *outlink = ctx->outputs[0]; - - mfxSyncPoint sync = NULL; - mfxStatus err; - - AVFrame *out = NULL; - int ret = 0; - - out = ff_get_video_buffer(outlink, outlink->w, outlink->h); - if (!out) { - ret = AVERROR(ENOMEM); - goto fail; - } + int ret = 0; + AVFilterContext *ctx = link->dst; + QSVVPPContext *qsv = ctx->priv; - do { - err = MFXVideoVPP_RunFrameVPPAsync(s->session, - (mfxFrameSurface1*)in->data[3], - (mfxFrameSurface1*)out->data[3], - NULL, &sync); - if (err == MFX_WRN_DEVICE_BUSY) - av_usleep(1); - } while (err == MFX_WRN_DEVICE_BUSY); - - if (err < 0) { - ret = ff_qsvvpp_print_error(ctx, err, "Error during scaling"); - goto fail; - } - - if (!sync) { - av_log(ctx, AV_LOG_ERROR, "No sync during scaling\n"); - ret = AVERROR_UNKNOWN; - goto fail; - } - - do { - err = MFXVideoCORE_SyncOperation(s->session, sync, 1000); - } while (err == MFX_WRN_IN_EXECUTION); - if (err < 0) { - ret = ff_qsvvpp_print_error(ctx, err, "Error synchronizing the operation"); - goto fail; - } - - ret = av_frame_copy_props(out, in); - if (ret < 0) - goto fail; - - out->width = outlink->w; - out->height = outlink->h; - - av_reduce(&out->sample_aspect_ratio.num, &out->sample_aspect_ratio.den, - (int64_t)in->sample_aspect_ratio.num * outlink->h * link->w, - (int64_t)in->sample_aspect_ratio.den * outlink->w * link->h, - INT_MAX); - - av_frame_free(&in); - return ff_filter_frame(outlink, out); -fail: + ret = ff_qsvvpp_filter_frame(qsv, link, in); av_frame_free(&in); - av_frame_free(&out); return ret; }

1 0

libavcodec/qsvenc: enable Hyper Encode
by galinart 30 Jan '23

30 Jan '23

ffmpeg | branch: master | galinart <artem.galin(a)gmail.com> | Mon Dec 5 20:18:21 2022 +0000| [500282941655558e2440afe163f0268dc5ac61bf] | committer: Haihao Xiang libavcodec/qsvenc: enable Hyper Encode Hyper Encode uses Intel integrated and discrete graphics on one system to accelerate encoding of a single video stream. Depending on the selected parameters and codecs, performance gain on AlderLake iGPU + ARC Gfx up to 1.6x. More information: https://www.intel.co.uk/content/www/uk/en/architecture-and-technology/adapt… Developer guide: https://github.com/oneapi-src/oneVPL-intel-gpu/blob/main/doc/HyperEncode_Fe… Hyper Encode is supported only on Windows and requires D3D11 and oneVPL. To enable Hyper Encode need to specify: -Hyper Encode mode (-dual_gfx on or dual_gfx adaptive) -Encoder: h264_qsv or hevc_qsv -BRC: VBR, CQP or ICQ -Lowpower mode (-low_power 1) -Closed GOP for AVC or strict GOP for HEVC -idr_interval = 0 used by default Depending on the encoding parameters, the following parameters may need to be adjusted: -g recommended >= 30 for better performance -async_depth recommended >= 30 for better performance -extra_hw_frames recommended equal to async_depth value -bf recommended = 0 for better performance In the cases with fast encoding (-preset veryfast) there may be no performance gain due to the fact that the decode is slower than the encode. Command line examples: ffmpeg.exe -init_hw_device qsv:hw,child_device_type=d3d11va,child_device=0 -v verbose -y -hwaccel qsv -extra_hw_frames 60 -async_depth 60 -c:v h264_qsv -i bbb_sunflower_2160p_60fps_normal.mp4 -async_depth 60 -c:v h264_qsv -preset medium -g 60 -low_power 1 -bf 0 -dual_gfx on output.h265 Signed-off-by: galinart <artem.galin(a)intel.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=500282941655558e2… --- libavcodec/qsvenc.c | 91 ++++++++++++++++++++++++++++++++++++++++++++++-- libavcodec/qsvenc.h | 19 ++++++++-- libavcodec/qsvenc_h264.c | 3 ++ libavcodec/qsvenc_hevc.c | 3 ++ 4 files changed, 112 insertions(+), 4 deletions(-) diff --git a/libavcodec/qsvenc.c b/libavcodec/qsvenc.c index 93f1862a4b..2f0e94a914 100644 --- a/libavcodec/qsvenc.c +++ b/libavcodec/qsvenc.c @@ -169,6 +169,8 @@ do { \ } \ } while (0) \ +#define MFX_IMPL_VIA_MASK(impl) (0x0f00 & (impl)) + static const char *print_ratecontrol(mfxU16 rc_mode) { int i; @@ -197,6 +199,10 @@ static void dump_video_param(AVCodecContext *avctx, QSVEncContext *q, mfxExtCodingOption2 *co2 = NULL; mfxExtCodingOption3 *co3 = NULL; mfxExtHEVCTiles *exthevctiles = NULL; +#if QSV_HAVE_HE + mfxExtHyperModeParam *exthypermodeparam = NULL; +#endif + const char *tmp_str = NULL; if (q->co2_idx > 0) @@ -208,6 +214,11 @@ static void dump_video_param(AVCodecContext *avctx, QSVEncContext *q, if (q->exthevctiles_idx > 0) exthevctiles = (mfxExtHEVCTiles *)coding_opts[q->exthevctiles_idx]; +#if QSV_HAVE_HE + if (q->exthypermodeparam_idx > 0) + exthypermodeparam = (mfxExtHyperModeParam *)coding_opts[q->exthypermodeparam_idx]; +#endif + av_log(avctx, AV_LOG_VERBOSE, "profile: %s; level: %"PRIu16"\n", print_profile(avctx->codec_id, info->CodecProfile), info->CodecLevel); @@ -373,6 +384,21 @@ static void dump_video_param(AVCodecContext *avctx, QSVEncContext *q, av_log(avctx, AV_LOG_VERBOSE, "NumTileColumns: %"PRIu16"; NumTileRows: %"PRIu16"\n", exthevctiles->NumTileColumns, exthevctiles->NumTileRows); } + +#if QSV_HAVE_HE + if (exthypermodeparam) { + av_log(avctx, AV_LOG_VERBOSE, "HyperEncode: "); + + if (exthypermodeparam->Mode == MFX_HYPERMODE_OFF) + av_log(avctx, AV_LOG_VERBOSE, "OFF"); + if (exthypermodeparam->Mode == MFX_HYPERMODE_ON) + av_log(avctx, AV_LOG_VERBOSE, "ON"); + if (exthypermodeparam->Mode == MFX_HYPERMODE_ADAPTIVE) + av_log(avctx, AV_LOG_VERBOSE, "Adaptive"); + + av_log(avctx, AV_LOG_VERBOSE, "\n"); + } +#endif } static void dump_video_vp9_param(AVCodecContext *avctx, QSVEncContext *q, @@ -1168,6 +1194,54 @@ static int init_video_param(AVCodecContext *avctx, QSVEncContext *q) q->extparam_internal[q->nb_extparam_internal++] = (mfxExtBuffer *)&q->extvsi; } +#if QSV_HAVE_HE + if (q->dual_gfx) { + if (QSV_RUNTIME_VERSION_ATLEAST(q->ver, 2, 4)) { + mfxIMPL impl; + MFXQueryIMPL(q->session, &impl); + + if (MFX_IMPL_VIA_MASK(impl) != MFX_IMPL_VIA_D3D11) { + av_log(avctx, AV_LOG_ERROR, "Dual GFX mode requires D3D11VA \n"); + return AVERROR_UNKNOWN; + } + if (q->param.mfx.LowPower != MFX_CODINGOPTION_ON) { + av_log(avctx, AV_LOG_ERROR, "Dual GFX mode supports only low-power encoding mode \n"); + return AVERROR_UNKNOWN; + } + if (q->param.mfx.CodecId != MFX_CODEC_AVC && q->param.mfx.CodecId != MFX_CODEC_HEVC) { + av_log(avctx, AV_LOG_ERROR, "Not supported encoder for dual GFX mode. " + "Supported: h264_qsv and hevc_qsv \n"); + return AVERROR_UNKNOWN; + } + if (q->param.mfx.RateControlMethod != MFX_RATECONTROL_VBR && + q->param.mfx.RateControlMethod != MFX_RATECONTROL_CQP && + q->param.mfx.RateControlMethod != MFX_RATECONTROL_ICQ) { + av_log(avctx, AV_LOG_WARNING, "Not supported BRC for dual GFX mode. " + "Supported: VBR, CQP and ICQ \n"); + } + if ((q->param.mfx.CodecId == MFX_CODEC_AVC && q->param.mfx.IdrInterval != 0) || + (q->param.mfx.CodecId == MFX_CODEC_HEVC && q->param.mfx.IdrInterval != 1)) { + av_log(avctx, AV_LOG_WARNING, "Dual GFX mode requires closed GOP for AVC and strict GOP for HEVC, -idr_interval 0 \n"); + } + if (q->param.mfx.GopPicSize < 30) { + av_log(avctx, AV_LOG_WARNING, "For better performance in dual GFX mode GopPicSize must be >= 30 \n"); + } + if (q->param.AsyncDepth < 30) { + av_log(avctx, AV_LOG_WARNING, "For better performance in dual GFX mode AsyncDepth must be >= 30 \n"); + } + + q->exthypermodeparam.Header.BufferId = MFX_EXTBUFF_HYPER_MODE_PARAM; + q->exthypermodeparam.Header.BufferSz = sizeof(q->exthypermodeparam); + q->exthypermodeparam.Mode = q->dual_gfx; + q->extparam_internal[q->nb_extparam_internal++] = (mfxExtBuffer *)&q->exthypermodeparam; + } else { + av_log(avctx, AV_LOG_ERROR, + "This version of runtime doesn't support Hyper Encode\n"); + return AVERROR_UNKNOWN; + } + } +#endif + if (!check_enc_param(avctx,q)) { av_log(avctx, AV_LOG_ERROR, "some encoding parameters are not supported by the QSV " @@ -1342,12 +1416,19 @@ static int qsv_retrieve_enc_params(AVCodecContext *avctx, QSVEncContext *q) .Header.BufferSz = sizeof(hevc_tile_buf), }; - mfxExtBuffer *ext_buffers[6]; +#if QSV_HAVE_HE + mfxExtHyperModeParam hyper_mode_param_buf = { + .Header.BufferId = MFX_EXTBUFF_HYPER_MODE_PARAM, + .Header.BufferSz = sizeof(hyper_mode_param_buf), + }; +#endif + + mfxExtBuffer *ext_buffers[6 + QSV_HAVE_HE]; int need_pps = avctx->codec_id != AV_CODEC_ID_MPEG2VIDEO; int ret, ext_buf_num = 0, extradata_offset = 0; - q->co2_idx = q->co3_idx = q->exthevctiles_idx = -1; + q->co2_idx = q->co3_idx = q->exthevctiles_idx = q->exthypermodeparam_idx = -1; ext_buffers[ext_buf_num++] = (mfxExtBuffer*)&extradata; ext_buffers[ext_buf_num++] = (mfxExtBuffer*)&co; @@ -1369,6 +1450,12 @@ static int qsv_retrieve_enc_params(AVCodecContext *avctx, QSVEncContext *q) q->exthevctiles_idx = ext_buf_num; ext_buffers[ext_buf_num++] = (mfxExtBuffer*)&hevc_tile_buf; } +#if QSV_HAVE_HE + if (q->dual_gfx && QSV_RUNTIME_VERSION_ATLEAST(q->ver, 2, 4)) { + q->exthypermodeparam_idx = ext_buf_num; + ext_buffers[ext_buf_num++] = (mfxExtBuffer*)&hyper_mode_param_buf; + } +#endif q->param.ExtParam = ext_buffers; q->param.NumExtParam = ext_buf_num; diff --git a/libavcodec/qsvenc.h b/libavcodec/qsvenc.h index a7bbb3797e..4a6fa2caed 100644 --- a/libavcodec/qsvenc.h +++ b/libavcodec/qsvenc.h @@ -45,10 +45,12 @@ #define QSV_HAVE_AVBR 1 #define QSV_HAVE_VCM 1 #define QSV_HAVE_MF 0 +#define QSV_HAVE_HE QSV_VERSION_ATLEAST(2, 4) #else #define QSV_HAVE_AVBR 0 #define QSV_HAVE_VCM 0 #define QSV_HAVE_MF !QSV_ONEVPL +#define QSV_HAVE_HE 0 #endif #define QSV_COMMON_OPTS \ @@ -64,6 +66,14 @@ { "forced_idr", "Forcing I frames as IDR frames", OFFSET(qsv.forced_idr), AV_OPT_TYPE_BOOL,{ .i64 = 0 }, 0, 1, VE }, \ { "low_power", "enable low power mode(experimental: many limitations by mfx version, BRC modes, etc.)", OFFSET(qsv.low_power), AV_OPT_TYPE_BOOL, { .i64 = -1}, -1, 1, VE}, +#if QSV_HAVE_HE +#define QSV_HE_OPTIONS \ +{ "dual_gfx", "Prefer processing on both iGfx and dGfx simultaneously", OFFSET(qsv.dual_gfx), AV_OPT_TYPE_INT, { .i64 = MFX_HYPERMODE_OFF }, MFX_HYPERMODE_OFF, MFX_HYPERMODE_ADAPTIVE, VE, "dual_gfx" }, \ +{ "off", "Disable HyperEncode mode", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_HYPERMODE_OFF }, INT_MIN, INT_MAX, VE, "dual_gfx" }, \ +{ "on", "Enable HyperEncode mode and return error if incompatible parameters during initialization", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_HYPERMODE_ON }, INT_MIN, INT_MAX, VE, "dual_gfx" }, \ +{ "adaptive", "Enable HyperEncode mode or fallback to single GPU if incompatible parameters during initialization", 0, AV_OPT_TYPE_CONST, { .i64 = MFX_HYPERMODE_ADAPTIVE }, INT_MIN, INT_MAX, VE, "dual_gfx" }, +#endif + #define QSV_OPTION_RDO \ { "rdo", "Enable rate distortion optimization", OFFSET(qsv.rdo), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 1, VE }, @@ -171,7 +181,9 @@ typedef struct QSVEncContext { mfxExtAV1TileParam extav1tileparam; mfxExtAV1BitstreamParam extav1bsparam; #endif - +#if QSV_HAVE_HE + mfxExtHyperModeParam exthypermodeparam; +#endif #if QSV_HAVE_OPAQUE mfxExtOpaqueSurfaceAlloc opaque_alloc; mfxFrameSurface1 **opaque_surfaces; @@ -180,7 +192,7 @@ typedef struct QSVEncContext { mfxExtVideoSignalInfo extvsi; - mfxExtBuffer *extparam_internal[5 + (QSV_HAVE_MF * 2) + QSV_HAVE_EXT_AV1_PARAM * 2]; + mfxExtBuffer *extparam_internal[5 + (QSV_HAVE_MF * 2) + (QSV_HAVE_EXT_AV1_PARAM * 2) + QSV_HAVE_HE]; int nb_extparam_internal; mfxExtBuffer **extparam; @@ -255,6 +267,7 @@ typedef struct QSVEncContext { int co2_idx; int co3_idx; int exthevctiles_idx; + int exthypermodeparam_idx; int vp9_idx; int max_qp_i; @@ -299,6 +312,8 @@ typedef struct QSVEncContext { // This is used for SEI Timing reset int old_pic_timing_sei; int skip_frame; + // This is used for Hyper Encode + int dual_gfx; } QSVEncContext; int ff_qsv_enc_init(AVCodecContext *avctx, QSVEncContext *q); diff --git a/libavcodec/qsvenc_h264.c b/libavcodec/qsvenc_h264.c index bbdedc3335..071a9a79e9 100644 --- a/libavcodec/qsvenc_h264.c +++ b/libavcodec/qsvenc_h264.c @@ -117,6 +117,9 @@ static const AVOption options[] = { QSV_OPTION_SCENARIO QSV_OPTION_AVBR QSV_OPTION_SKIP_FRAME +#if QSV_HAVE_HE + QSV_HE_OPTIONS +#endif { "cavlc", "Enable CAVLC", OFFSET(qsv.cavlc), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE }, #if QSV_HAVE_VCM diff --git a/libavcodec/qsvenc_hevc.c b/libavcodec/qsvenc_hevc.c index e042263bf5..5e23ca9647 100644 --- a/libavcodec/qsvenc_hevc.c +++ b/libavcodec/qsvenc_hevc.c @@ -318,6 +318,9 @@ static const AVOption options[] = { QSV_OPTION_SCENARIO QSV_OPTION_AVBR QSV_OPTION_SKIP_FRAME +#if QSV_HAVE_HE + QSV_HE_OPTIONS +#endif { "idr_interval", "Distance (in I-frames) between IDR frames", OFFSET(qsv.idr_interval), AV_OPT_TYPE_INT, { .i64 = 0 }, -1, INT_MAX, VE, "idr_interval" }, { "begin_only", "Output an IDR-frame only at the beginning of the stream", 0, AV_OPT_TYPE_CONST, { .i64 = -1 }, 0, 0, VE, "idr_interval" },

1 0

avfilter: add VA-API variants of the stack filters
by Haihao Xiang 30 Jan '23

30 Jan '23

ffmpeg | branch: master | Haihao Xiang <haihao.xiang(a)intel.com> | Thu Jan 19 16:30:06 2023 +0800| [aecfec6f80fd8be68e050a97935589beeae588fa] | committer: Haihao Xiang avfilter: add VA-API variants of the stack filters Include hstack_vaapi, vstack_vaapi and xstack_vaapi. They may accept input streams with different sizes. libva2 (VA-API 1.0+) is required. Example: $ ffmpeg -hwaccel vaapi -hwaccel_output_format vaapi -i input.h265 -filter_complex "[0:v][0:v]hstack_vaapi" -c:v h264_vaapi out.h264 $ ffmpeg \ -hwaccel vaapi -hwaccel_output_format vaapi -i input.h264 \ -hwaccel vaapi -hwaccel_output_format vaapi -i input.h264 \ -hwaccel vaapi -hwaccel_output_format vaapi -i input.h264 \ -hwaccel vaapi -hwaccel_output_format vaapi -i input.h264 \ -filter_complex "[0:v][1:v][2:v][3:v]xstack_vaapi=inputs=4:fill=0x000000:layout=0_0_1920x1080|w0_0_1920x1080|0_h0_1920x1080|w0_h0_1920x1080" \ -c:v hevc_vaapi out.h265 Signed-off-by: Haihao Xiang <haihao.xiang(a)intel.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=aecfec6f80fd8be68… --- Changelog | 1 + configure | 3 + doc/filters.texi | 81 ++++++ libavfilter/Makefile | 3 + libavfilter/allfilters.c | 3 + libavfilter/version.h | 2 +- libavfilter/vf_stack_vaapi.c | 574 +++++++++++++++++++++++++++++++++++++++++++ 7 files changed, 666 insertions(+), 1 deletion(-) diff --git a/Changelog b/Changelog index e07a039ca4..a0f1ad7211 100644 --- a/Changelog +++ b/Changelog @@ -33,6 +33,7 @@ version <next>: - CBD2 DPCM decoder - ssim360 video filter - ffmpeg CLI new options: -enc_stats_pre[_fmt], -enc_stats_post[_fmt] +- hstack_vaapi, vstack_vaapi and xstack_vaapi filters version 5.1: diff --git a/configure b/configure index cb895cf9ac..8190e978a8 100755 --- a/configure +++ b/configure @@ -3766,6 +3766,9 @@ xfade_opencl_filter_deps="opencl" yadif_cuda_filter_deps="ffnvcodec" yadif_cuda_filter_deps_any="cuda_nvcc cuda_llvm" yadif_videotoolbox_filter_deps="metal corevideo videotoolbox" +hstack_vaapi_filter_deps="vaapi_1" +vstack_vaapi_filter_deps="vaapi_1" +xstack_vaapi_filter_deps="vaapi_1" # examples avio_list_dir_deps="avformat avutil" diff --git a/doc/filters.texi b/doc/filters.texi index 4f4d2ea807..3a54c68f3e 100644 --- a/doc/filters.texi +++ b/doc/filters.texi @@ -14941,6 +14941,7 @@ Set the scaling dimension: @code{2} for @code{hq2x}, @code{3} for Default is @code{3}. @end table +@anchor{hstack} @section hstack Stack input videos horizontally. @@ -24392,6 +24393,7 @@ Example: ffmpeg -i ref.mpg -vf vmafmotion -f null - @end example +@anchor{vstack} @section vstack Stack input videos vertically. @@ -24873,6 +24875,7 @@ minimum values, and @code{1} maximum values. This filter supports all above options as @ref{commands}, excluding option @code{inputs}. +@anchor{xstack} @section xstack Stack video inputs into custom layout. @@ -26689,6 +26692,84 @@ tonemap_vaapi=format=p010:t=bt2020-10 @end example @end itemize +@section hstack_vaapi +Stack input videos horizontally. + +This is the VA-API variant of the @ref{hstack} filter, each input stream may +have different height, this filter will scale down/up each input stream while +keeping the orignal aspect. + +It accepts the following options: + +@table @option +@item inputs +See @ref{hstack}. + +@item shortest +See @ref{hstack}. + +@item height +Set height of output. If set to 0, this filter will set height of output to +height of the first input stream. Default value is 0. +@end table + +@section vstack_vaapi +Stack input videos vertically. + +This is the VA-API variant of the @ref{vstack} filter, each input stream may +have different width, this filter will scale down/up each input stream while +keeping the orignal aspect. + +It accepts the following options: + +@table @option +@item inputs +See @ref{vstack}. + +@item shortest +See @ref{vstack}. + +@item width +Set width of output. If set to 0, this filter will set width of output to +width of the first input stream. Default value is 0. +@end table + +@section xstack_vaapi +Stack video inputs into custom layout. + +This is the VA-API variant of the @ref{xstack} filter, each input stream may +have different size, this filter will scale down/up each input stream to the +given output size, or the size of the first input stream. + +It accepts the following options: + +@table @option +@item inputs +See @ref{xstack}. + +@item shortest +See @ref{xstack}. + +@item layout +See @ref{xstack}. +Moreover, this permits the user to supply output size for each input stream. +@example +xstack_vaapi=inputs=4:layout=0_0_1920x1080|0_h0_1920x1080|w0_0_1920x1080|w0_h0_1920x1080 +@end example + +@item grid +See @ref{xstack}. + +@item grid_tile_size +Set output size for each input stream when @option{grid} is set. If this option +is not set, this filter will set output size by default to the size of the +first input stream. For the syntax of this option, check the +@ref{video size syntax,,"Video size" section in the ffmpeg-utils manual,ffmpeg-utils}. + +@item fill +See @ref{xstack}. +@end table + @c man end VAAPI VIDEO FILTERS @chapter Video Sources diff --git a/libavfilter/Makefile b/libavfilter/Makefile index d4e38bd4e8..a0059c69e6 100644 --- a/libavfilter/Makefile +++ b/libavfilter/Makefile @@ -558,6 +558,9 @@ OBJS-$(CONFIG_YAEPBLUR_FILTER) += vf_yaepblur.o OBJS-$(CONFIG_ZMQ_FILTER) += f_zmq.o OBJS-$(CONFIG_ZOOMPAN_FILTER) += vf_zoompan.o OBJS-$(CONFIG_ZSCALE_FILTER) += vf_zscale.o +OBJS-$(CONFIG_HSTACK_VAAPI_FILTER) += vf_stack_vaapi.o framesync.o vaapi_vpp.o +OBJS-$(CONFIG_VSTACK_VAAPI_FILTER) += vf_stack_vaapi.o framesync.o vaapi_vpp.o +OBJS-$(CONFIG_XSTACK_VAAPI_FILTER) += vf_stack_vaapi.o framesync.o vaapi_vpp.o OBJS-$(CONFIG_ALLRGB_FILTER) += vsrc_testsrc.o OBJS-$(CONFIG_ALLYUV_FILTER) += vsrc_testsrc.o diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c index 5c86d0789f..9cdcca4853 100644 --- a/libavfilter/allfilters.c +++ b/libavfilter/allfilters.c @@ -523,6 +523,9 @@ extern const AVFilter ff_vf_yaepblur; extern const AVFilter ff_vf_zmq; extern const AVFilter ff_vf_zoompan; extern const AVFilter ff_vf_zscale; +extern const AVFilter ff_vf_hstack_vaapi; +extern const AVFilter ff_vf_vstack_vaapi; +extern const AVFilter ff_vf_xstack_vaapi; extern const AVFilter ff_vsrc_allrgb; extern const AVFilter ff_vsrc_allyuv; diff --git a/libavfilter/version.h b/libavfilter/version.h index 09b9954597..057ab63415 100644 --- a/libavfilter/version.h +++ b/libavfilter/version.h @@ -31,7 +31,7 @@ #include "version_major.h" -#define LIBAVFILTER_VERSION_MINOR 55 +#define LIBAVFILTER_VERSION_MINOR 56 #define LIBAVFILTER_VERSION_MICRO 100 diff --git a/libavfilter/vf_stack_vaapi.c b/libavfilter/vf_stack_vaapi.c new file mode 100644 index 0000000000..403fbb19eb --- /dev/null +++ b/libavfilter/vf_stack_vaapi.c @@ -0,0 +1,574 @@ +/* + * 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 + */ + +/** + * @file + * Hardware accelerated hstack, vstack and xstack filters based on VA-API + */ + +#include "config_components.h" + +#include "libavutil/opt.h" +#include "libavutil/common.h" +#include "libavutil/pixdesc.h" +#include "libavutil/eval.h" +#include "libavutil/hwcontext.h" +#include "libavutil/avstring.h" +#include "libavutil/avassert.h" +#include "libavutil/imgutils.h" +#include "libavutil/mathematics.h" +#include "libavutil/parseutils.h" +#include "libavutil/mem.h" + +#include "internal.h" +#include "filters.h" +#include "formats.h" +#include "video.h" +#include "framesync.h" +#include "vaapi_vpp.h" + +#define OFFSET(x) offsetof(StackVAAPIContext, x) +#define FLAGS (AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_FILTERING_PARAM) + +enum { + STACK_VAAPI_H = 0, + STACK_VAAPI_V = 1, + STACK_VAAPI_X = 2 +}; + +typedef struct StackVAAPIContext { + VAAPIVPPContext vppctx; /**< must be the first field */ + + FFFrameSync fs; + int mode; + VARectangle *rects; + uint8_t fillcolor[4]; + int fillcolor_enable; + + /* Options */ + int nb_inputs; + int shortest; + int tile_width; + int tile_height; + int nb_grid_columns; + int nb_grid_rows; + char *layout; + char *fillcolor_str; +} StackVAAPIContext; + +static int process_frame(FFFrameSync *fs) +{ + AVFilterContext *avctx = fs->parent; + AVFilterLink *outlink = avctx->outputs[0]; + StackVAAPIContext *sctx = fs->opaque; + VAAPIVPPContext *vppctx = fs->opaque; + AVFrame *oframe, *iframe; + VAProcPipelineParameterBuffer *params = NULL; + VARectangle *irect = NULL; + int ret = 0; + + if (vppctx->va_context == VA_INVALID_ID) + return AVERROR(EINVAL); + + oframe = ff_get_video_buffer(outlink, outlink->w, outlink->h); + if (!oframe) + return AVERROR(ENOMEM); + + irect = av_calloc(avctx->nb_inputs, sizeof(*irect)); + params = av_calloc(avctx->nb_inputs, sizeof(*params)); + if (!irect || !params) { + ret = AVERROR(ENOMEM); + goto fail; + } + + for (int i = 0; i < avctx->nb_inputs; i++) { + ret = ff_framesync_get_frame(fs, i, &iframe, 0); + if (ret) + goto fail; + + if (i == 0) { + ret = av_frame_copy_props(oframe, iframe); + if (ret < 0) + goto fail; + } + + ret = ff_vaapi_vpp_init_params(avctx, &params[i], iframe, oframe); + if (ret) + goto fail; + + av_log(avctx, AV_LOG_DEBUG, "stack input %d: %s, %ux%u (%"PRId64").\n", + i, av_get_pix_fmt_name(iframe->format), + iframe->width, iframe->height, iframe->pts); + irect[i].x = 0; + irect[i].y = 0; + irect[i].width = iframe->width; + irect[i].height = iframe->height; + params[i].surface_region = &irect[i]; + params[i].surface = (VASurfaceID)(uintptr_t)iframe->data[3]; + params[i].output_region = &sctx->rects[i]; + + if (sctx->fillcolor_enable) + params[i].output_background_color = (sctx->fillcolor[3] << 24 | + sctx->fillcolor[0] << 16 | + sctx->fillcolor[1] << 8 | + sctx->fillcolor[2]); + } + + oframe->pts = av_rescale_q(sctx->fs.pts, sctx->fs.time_base, outlink->time_base); + oframe->sample_aspect_ratio = outlink->sample_aspect_ratio; + + ret = ff_vaapi_vpp_render_pictures(avctx, params, avctx->nb_inputs, oframe); + if (ret) + goto fail; + + av_freep(&irect); + av_freep(&params); + return ff_filter_frame(outlink, oframe); + +fail: + av_freep(&irect); + av_freep(&params); + av_frame_free(&oframe); + return ret; +} + +static int init_framesync(AVFilterContext *avctx) +{ + StackVAAPIContext *sctx = avctx->priv; + int ret; + + ret = ff_framesync_init(&sctx->fs, avctx, avctx->nb_inputs); + if (ret < 0) + return ret; + + sctx->fs.on_event = process_frame; + sctx->fs.opaque = sctx; + + for (int i = 0; i < sctx->nb_inputs; i++) { + FFFrameSyncIn *in = &sctx->fs.in[i]; + + in->before = EXT_STOP; + in->after = sctx->shortest ? EXT_STOP : EXT_INFINITY; + in->sync = 1; + in->time_base = avctx->inputs[i]->time_base; + } + + return ff_framesync_configure(&sctx->fs); +} + +#define SET_INPUT_REGION(rect, rx, ry, rw, rh) do { \ + rect->x = rx; \ + rect->y = ry; \ + rect->width = rw; \ + rect->height = rh; \ + } while (0) + +static int config_output(AVFilterLink *outlink) +{ + AVFilterContext *avctx = outlink->src; + StackVAAPIContext *sctx = avctx->priv; + VAAPIVPPContext *vppctx = avctx->priv; + AVFilterLink *inlink0 = avctx->inputs[0]; + AVHWFramesContext *hwfc0 = NULL; + int width, height, ret; + + if (inlink0->format != AV_PIX_FMT_VAAPI || !inlink0->hw_frames_ctx || !inlink0->hw_frames_ctx->data) { + av_log(avctx, AV_LOG_ERROR, "Software pixel format is not supported.\n"); + return AVERROR(EINVAL); + } + + hwfc0 = (AVHWFramesContext *)inlink0->hw_frames_ctx->data; + + for (int i = 1; i < sctx->nb_inputs; i++) { + AVFilterLink *inlink = avctx->inputs[i]; + AVHWFramesContext *hwfc = NULL; + + if (inlink->format != AV_PIX_FMT_VAAPI || !inlink->hw_frames_ctx || !inlink->hw_frames_ctx->data) { + av_log(avctx, AV_LOG_ERROR, "Software pixel format is not supported.\n"); + return AVERROR(EINVAL); + } + + hwfc = (AVHWFramesContext *)inlink->hw_frames_ctx->data; + + if (hwfc0->sw_format != hwfc->sw_format) { + av_log(avctx, AV_LOG_ERROR, "All inputs should have the same underlying software pixel format.\n"); + return AVERROR(EINVAL); + } + + if (hwfc0->device_ctx != hwfc->device_ctx) { + av_log(avctx, AV_LOG_ERROR, "All inputs should have the same underlying vaapi devices.\n"); + return AVERROR(EINVAL); + } + } + + ff_vaapi_vpp_config_input(inlink0); + vppctx->output_format = hwfc0->sw_format; + + if (sctx->mode == STACK_VAAPI_H) { + height = sctx->tile_height; + width = 0; + + if (!height) + height = inlink0->h; + + for (int i = 0; i < sctx->nb_inputs; i++) { + AVFilterLink *inlink = avctx->inputs[i]; + VARectangle *rect = &sctx->rects[i]; + + SET_INPUT_REGION(rect, width, 0, av_rescale(height, inlink->w, inlink->h), height); + width += av_rescale(height, inlink->w, inlink->h); + } + } else if (sctx->mode == STACK_VAAPI_V) { + height = 0; + width = sctx->tile_width; + + if (!width) + width = inlink0->w; + + for (int i = 0; i < sctx->nb_inputs; i++) { + AVFilterLink *inlink = avctx->inputs[i]; + VARectangle *rect = &sctx->rects[i]; + + SET_INPUT_REGION(rect, 0, height, width, av_rescale(width, inlink->h, inlink->w)); + height += av_rescale(width, inlink->h, inlink->w); + } + } else if (sctx->nb_grid_rows && sctx->nb_grid_columns) { + int xpos = 0, ypos = 0; + int ow, oh, k = 0; + + ow = sctx->tile_width; + oh = sctx->tile_height; + + if (!ow || !oh) { + ow = avctx->inputs[0]->w; + oh = avctx->inputs[0]->h; + } + + for (int i = 0; i < sctx->nb_grid_columns; i++) { + ypos = 0; + + for (int j = 0; j < sctx->nb_grid_rows; j++) { + VARectangle *rect = &sctx->rects[k++]; + + SET_INPUT_REGION(rect, xpos, ypos, ow, oh); + ypos += oh; + } + + xpos += ow; + } + + width = ow * sctx->nb_grid_columns; + height = oh * sctx->nb_grid_rows; + } else { + char *arg, *p = sctx->layout, *saveptr = NULL; + char *arg2, *p2, *saveptr2 = NULL; + char *arg3, *p3, *saveptr3 = NULL; + int xpos, ypos, size; + int ow, oh; + + width = avctx->inputs[0]->w; + height = avctx->inputs[0]->h; + + for (int i = 0; i < sctx->nb_inputs; i++) { + AVFilterLink *inlink = avctx->inputs[i]; + VARectangle *rect = &sctx->rects[i]; + + ow = inlink->w; + oh = inlink->h; + + if (!(arg = av_strtok(p, "|", &saveptr))) + return AVERROR(EINVAL); + + p = NULL; + p2 = arg; + xpos = ypos = 0; + + for (int j = 0; j < 3; j++) { + if (!(arg2 = av_strtok(p2, "_", &saveptr2))) { + if (j == 2) + break; + else + return AVERROR(EINVAL); + } + + p2 = NULL; + p3 = arg2; + + if (j == 2) { + if ((ret = av_parse_video_size(&ow, &oh, p3)) < 0) { + av_log(avctx, AV_LOG_ERROR, "Invalid size '%s'\n", p3); + return ret; + } + + break; + } + + while ((arg3 = av_strtok(p3, "+", &saveptr3))) { + p3 = NULL; + if (sscanf(arg3, "w%d", &size) == 1) { + if (size == i || size < 0 || size >= sctx->nb_inputs) + return AVERROR(EINVAL); + + if (!j) + xpos += sctx->rects[size].width; + else + ypos += sctx->rects[size].width; + } else if (sscanf(arg3, "h%d", &size) == 1) { + if (size == i || size < 0 || size >= sctx->nb_inputs) + return AVERROR(EINVAL); + + if (!j) + xpos += sctx->rects[size].height; + else + ypos += sctx->rects[size].height; + } else if (sscanf(arg3, "%d", &size) == 1) { + if (size < 0) + return AVERROR(EINVAL); + + if (!j) + xpos += size; + else + ypos += size; + } else { + return AVERROR(EINVAL); + } + } + } + + SET_INPUT_REGION(rect, xpos, ypos, ow, oh); + width = FFMAX(width, xpos + ow); + height = FFMAX(height, ypos + oh); + } + + } + + outlink->w = width; + outlink->h = height; + outlink->frame_rate = inlink0->frame_rate; + outlink->sample_aspect_ratio = inlink0->sample_aspect_ratio; + + for (int i = 1; i < sctx->nb_inputs; i++) { + AVFilterLink *inlink = avctx->inputs[i]; + if (outlink->frame_rate.num != inlink->frame_rate.num || + outlink->frame_rate.den != inlink->frame_rate.den) { + av_log(avctx, AV_LOG_VERBOSE, + "Video inputs have different frame rates, output will be VFR\n"); + outlink->frame_rate = av_make_q(1, 0); + break; + } + } + + ret = init_framesync(avctx); + if (ret < 0) + return ret; + + outlink->time_base = sctx->fs.time_base; + + vppctx->output_width = width; + vppctx->output_height = height; + + return ff_vaapi_vpp_config_output(outlink); +} + +static int vaapi_stack_init(AVFilterContext *avctx) +{ + StackVAAPIContext *sctx = avctx->priv; + VAAPIVPPContext *vppctx = avctx->priv; + int ret; + + if (!strcmp(avctx->filter->name, "hstack_vaapi")) + sctx->mode = STACK_VAAPI_H; + else if (!strcmp(avctx->filter->name, "vstack_vaapi")) + sctx->mode = STACK_VAAPI_V; + else { + int is_grid; + + av_assert0(strcmp(avctx->filter->name, "xstack_vaapi") == 0); + sctx->mode = STACK_VAAPI_X; + is_grid = sctx->nb_grid_rows && sctx->nb_grid_columns; + + if (sctx->layout && is_grid) { + av_log(avctx, AV_LOG_ERROR, "Both layout and grid were specified. Only one is allowed.\n"); + return AVERROR(EINVAL); + } + + if (!sctx->layout && !is_grid) { + if (sctx->nb_inputs == 2) { + sctx->nb_grid_rows = 1; + sctx->nb_grid_columns = 2; + is_grid = 1; + } else { + av_log(avctx, AV_LOG_ERROR, "No layout or grid specified.\n"); + return AVERROR(EINVAL); + } + } + + if (is_grid) + sctx->nb_inputs = sctx->nb_grid_rows * sctx->nb_grid_columns; + + if (strcmp(sctx->fillcolor_str, "none") && + av_parse_color(sctx->fillcolor, sctx->fillcolor_str, -1, avctx) >= 0) { + sctx->fillcolor_enable = 1; + } else { + sctx->fillcolor_enable = 0; + } + } + + for (int i = 0; i < sctx->nb_inputs; i++) { + AVFilterPad pad = { 0 }; + + pad.type = AVMEDIA_TYPE_VIDEO; + pad.name = av_asprintf("input%d", i); + + if (!pad.name) + return AVERROR(ENOMEM); + + if ((ret = ff_append_inpad_free_name(avctx, &pad)) < 0) + return ret; + } + + /* stack region */ + sctx->rects = av_calloc(sctx->nb_inputs, sizeof(*sctx->rects)); + if (!sctx->rects) + return AVERROR(ENOMEM); + + ff_vaapi_vpp_ctx_init(avctx); + vppctx->output_format = AV_PIX_FMT_NONE; + + return 0; +} + +static av_cold void vaapi_stack_uninit(AVFilterContext *avctx) +{ + StackVAAPIContext *sctx = avctx->priv; + + ff_framesync_uninit(&sctx->fs); + av_freep(&sctx->rects); +} + +static int vaapi_stack_activate(AVFilterContext *avctx) +{ + StackVAAPIContext *sctx = avctx->priv; + return ff_framesync_activate(&sctx->fs); +} + +static int vaapi_stack_query_formats(AVFilterContext *avctx) +{ + static const enum AVPixelFormat pixel_formats[] = { + AV_PIX_FMT_VAAPI, + AV_PIX_FMT_NONE, + }; + + return ff_set_common_formats_from_list(avctx, pixel_formats); +} + +static const AVFilterPad vaapi_stack_outputs[] = { + { + .name = "default", + .type = AVMEDIA_TYPE_VIDEO, + .config_props = config_output, + }, +}; + +#define STACK_COMMON_OPTS \ + { "inputs", "Set number of inputs", OFFSET(nb_inputs), AV_OPT_TYPE_INT, { .i64 = 2 }, 2, UINT16_MAX, .flags = FLAGS }, \ + { "shortest", "Force termination when the shortest input terminates", OFFSET(shortest), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, FLAGS }, + +#if CONFIG_HSTACK_VAAPI_FILTER + +static const AVOption hstack_vaapi_options[] = { + STACK_COMMON_OPTS + + { "height", "Set output height (0 to use the height of input 0)", OFFSET(tile_height), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, UINT16_MAX, FLAGS }, + { NULL } +}; + +AVFILTER_DEFINE_CLASS(hstack_vaapi); + +const AVFilter ff_vf_hstack_vaapi = { + .name = "hstack_vaapi", + .description = NULL_IF_CONFIG_SMALL("VA-API hstack."), + .priv_size = sizeof(StackVAAPIContext), + .priv_class = &hstack_vaapi_class, + .init = vaapi_stack_init, + .uninit = vaapi_stack_uninit, + .activate = vaapi_stack_activate, + FILTER_QUERY_FUNC(vaapi_stack_query_formats), + FILTER_OUTPUTS(vaapi_stack_outputs), + .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE, + .flags = AVFILTER_FLAG_DYNAMIC_INPUTS, +}; + +#endif + +#if CONFIG_VSTACK_VAAPI_FILTER + +static const AVOption vstack_vaapi_options[] = { + STACK_COMMON_OPTS + + { "width", "Set output width (0 to use the width of input 0)", OFFSET(tile_width), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, UINT16_MAX, FLAGS }, + { NULL } +}; + +AVFILTER_DEFINE_CLASS(vstack_vaapi); + +const AVFilter ff_vf_vstack_vaapi = { + .name = "vstack_vaapi", + .description = NULL_IF_CONFIG_SMALL("VA-API vstack."), + .priv_size = sizeof(StackVAAPIContext), + .priv_class = &vstack_vaapi_class, + .init = vaapi_stack_init, + .uninit = vaapi_stack_uninit, + .activate = vaapi_stack_activate, + FILTER_QUERY_FUNC(vaapi_stack_query_formats), + FILTER_OUTPUTS(vaapi_stack_outputs), + .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE, + .flags = AVFILTER_FLAG_DYNAMIC_INPUTS, +}; + +#endif + +#if CONFIG_XSTACK_VAAPI_FILTER + +static const AVOption xstack_vaapi_options[] = { + STACK_COMMON_OPTS + + { "layout", "Set custom layout", OFFSET(layout), AV_OPT_TYPE_STRING, {.str = NULL}, 0, 0, .flags = FLAGS }, + { "grid", "set fixed size grid layout", OFFSET(nb_grid_columns), AV_OPT_TYPE_IMAGE_SIZE, {.str = NULL}, 0, 0, .flags = FLAGS }, + { "grid_tile_size", "set tile size in grid layout", OFFSET(tile_width), AV_OPT_TYPE_IMAGE_SIZE, {.str = NULL}, 0, 0, .flags = FLAGS }, + { "fill", "Set the color for unused pixels", OFFSET(fillcolor_str), AV_OPT_TYPE_STRING, {.str = "none"}, .flags = FLAGS }, + { NULL } +}; + +AVFILTER_DEFINE_CLASS(xstack_vaapi); + +const AVFilter ff_vf_xstack_vaapi = { + .name = "xstack_vaapi", + .description = NULL_IF_CONFIG_SMALL("VA-API xstack."), + .priv_size = sizeof(StackVAAPIContext), + .priv_class = &xstack_vaapi_class, + .init = vaapi_stack_init, + .uninit = vaapi_stack_uninit, + .activate = vaapi_stack_activate, + FILTER_OUTPUTS(vaapi_stack_outputs), + FILTER_QUERY_FUNC(vaapi_stack_query_formats), + .flags_internal = FF_FILTER_FLAG_HWFRAME_AWARE, + .flags = AVFILTER_FLAG_DYNAMIC_INPUTS, +}; + +#endif

1 0