April 2025 - ffmpeg-cvslog - lists.ffmpeg.org

avformat/hls: Fix Youtube AAC
by Michael Niedermayer 29 Apr '25

29 Apr '25

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Sun Apr 6 18:52:05 2025 +0200| [48c0dba23b3ce8c2bcb180bd2c8029c3c2875424] | committer: Michael Niedermayer avformat/hls: Fix Youtube AAC Fixes: Ticket11435 Fixes: yt-dlp -f 234+270 https://www.youtube.com/live/l8PMl7tUDIE Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=48c0dba23b3ce8c2b… --- libavformat/hls.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/libavformat/hls.c b/libavformat/hls.c index f2b98a6d72..84a5e66149 100644 --- a/libavformat/hls.c +++ b/libavformat/hls.c @@ -753,7 +753,8 @@ static int test_segment(AVFormatContext *s, const AVInputFormat *in_fmt, struct if (in_fmt->extensions) { matchF = av_match_ext( seg->url, in_fmt->extensions) + 2*(ff_match_url_ext(seg->url, in_fmt->extensions) > 0); - if(av_match_name("mp4", in_fmt->name)) { + // Youtube uses aac files with .ts extension + if(av_match_name("mp4", in_fmt->name) || av_match_name("aac", in_fmt->name)) { matchF |= av_match_ext( seg->url, "ts,m2t,m2ts,mts,mpg,m4s,mpeg,mpegts") + 2*(ff_match_url_ext(seg->url, "ts,m2t,m2ts,mts,mpg,m4s,mpeg,mpegts") > 0); }

1 0

tests: Add stream dump test API util, use it to dump stream data for chained ogg/{vorbis, opus, flac} streams.
by Romain Beauxis 29 Apr '25

29 Apr '25

ffmpeg | branch: master | Romain Beauxis <romain.beauxis(a)gmail.com> | Mon Apr 28 18:31:35 2025 -0500| [2431fd0b275869f055ca69db021270998e42362e] | committer: Michael Niedermayer tests: Add stream dump test API util, use it to dump stream data for chained ogg/{vorbis, opus, flac} streams. Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=2431fd0b275869f05… --- tests/Makefile | 4 + tests/api/Makefile | 2 +- tests/api/api-dump-stream-meta-test.c | 182 +++++++++++++++++++++++++++++ tests/fate/ogg-flac.mak | 11 ++ tests/fate/ogg-opus.mak | 11 ++ tests/fate/ogg-vorbis.mak | 11 ++ tests/ref/fate/ogg-flac-chained-meta.txt | 12 ++ tests/ref/fate/ogg-opus-chained-meta.txt | 27 +++++ tests/ref/fate/ogg-vorbis-chained-meta.txt | 17 +++ 9 files changed, 276 insertions(+), 1 deletion(-) diff --git a/tests/Makefile b/tests/Makefile index 0c08f68713..10871f28f8 100644 --- a/tests/Makefile +++ b/tests/Makefile @@ -220,6 +220,9 @@ include $(SRC_PATH)/tests/fate/mpeg4.mak include $(SRC_PATH)/tests/fate/mpegps.mak include $(SRC_PATH)/tests/fate/mpegts.mak include $(SRC_PATH)/tests/fate/mxf.mak +include $(SRC_PATH)/tests/fate/ogg-vorbis.mak +include $(SRC_PATH)/tests/fate/ogg-flac.mak +include $(SRC_PATH)/tests/fate/ogg-opus.mak include $(SRC_PATH)/tests/fate/oma.mak include $(SRC_PATH)/tests/fate/opus.mak include $(SRC_PATH)/tests/fate/pcm.mak @@ -278,6 +281,7 @@ $(FATE_FFPROBE) $(FATE_FFMPEG_FFPROBE) $(FATE_SAMPLES_FFPROBE) $(FATE_SAMPLES_FF $(FATE_SAMPLES_FASTSTART): tools/qt-faststart$(EXESUF) $(FATE_SAMPLES_DUMP_DATA) $(FATE_SAMPLES_DUMP_DATA-yes): tools/venc_data_dump$(EXESUF) $(FATE_SAMPLES_SCALE_SLICE): tools/scale_slice_test$(EXESUF) +$(FATE_SAMPLES_DUMP_STREAM_META): tests/api/api-dump-stream-meta-test$(EXESUF) ifdef SAMPLES FATE += $(FATE_EXTERN) diff --git a/tests/api/Makefile b/tests/api/Makefile index c96e636756..a2cb06a729 100644 --- a/tests/api/Makefile +++ b/tests/api/Makefile @@ -1,7 +1,7 @@ APITESTPROGS-$(call ENCDEC, FLAC, FLAC) += api-flac APITESTPROGS-$(call DEMDEC, H264, H264) += api-h264 APITESTPROGS-$(call DEMDEC, H264, H264) += api-h264-slice -APITESTPROGS-yes += api-seek +APITESTPROGS-yes += api-seek api-dump-stream-meta APITESTPROGS-$(call DEMDEC, H263, H263) += api-band APITESTPROGS-$(HAVE_THREADS) += api-threadmessage APITESTPROGS += $(APITESTPROGS-yes) diff --git a/tests/api/api-dump-stream-meta-test.c b/tests/api/api-dump-stream-meta-test.c new file mode 100644 index 0000000000..629b3a576a --- /dev/null +++ b/tests/api/api-dump-stream-meta-test.c @@ -0,0 +1,182 @@ +/* + * Copyright (c) 2025 Romain Beauxis + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to deal + * in the Software without restriction, including without limitation the rights + * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell + * copies of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN + * THE SOFTWARE. + */ + +/** + * Dump stream metadata + */ + +#include "libavcodec/avcodec.h" +#include "libavformat/avformat.h" +#include "libavutil/timestamp.h" + +static int dump_stream_meta(const char *input_filename) { + const AVCodec *codec = NULL; + AVPacket *pkt = NULL; + AVFrame *fr = NULL; + AVFormatContext *fmt_ctx = NULL; + AVCodecContext *ctx = NULL; + AVCodecParameters *origin_par = NULL; + AVStream *st; + int stream_idx = 0; + int result; + char *metadata; + + result = avformat_open_input(&fmt_ctx, input_filename, NULL, NULL); + if (result < 0) { + av_log(NULL, AV_LOG_ERROR, "Can't open file\n"); + return result; + } + + result = avformat_find_stream_info(fmt_ctx, NULL); + if (result < 0) { + av_log(NULL, AV_LOG_ERROR, "Can't get stream info\n"); + goto end; + } + + if (fmt_ctx->nb_streams > 1) { + av_log(NULL, AV_LOG_ERROR, "More than one stream found in input!\n"); + goto end; + } + + origin_par = fmt_ctx->streams[stream_idx]->codecpar; + st = fmt_ctx->streams[stream_idx]; + + result = av_dict_get_string(st->metadata, &metadata, '=', ':'); + if (result < 0) + goto end; + + printf("Stream ID: %d, codec name: %s, metadata: %s\n", stream_idx, + avcodec_get_name(origin_par->codec_id), + strlen(metadata) ? metadata : "N/A"); + + codec = avcodec_find_decoder(origin_par->codec_id); + if (!codec) { + av_log(NULL, AV_LOG_ERROR, "Can't find decoder\n"); + result = AVERROR_DECODER_NOT_FOUND; + goto end; + } + + ctx = avcodec_alloc_context3(codec); + if (!ctx) { + av_log(NULL, AV_LOG_ERROR, "Can't allocate decoder context\n"); + result = AVERROR(ENOMEM); + goto end; + } + + result = avcodec_parameters_to_context(ctx, origin_par); + if (result) { + av_log(NULL, AV_LOG_ERROR, "Can't copy decoder context\n"); + goto end; + } + + result = avcodec_open2(ctx, codec, NULL); + if (result < 0) { + av_log(ctx, AV_LOG_ERROR, "Can't open decoder\n"); + goto end; + } + + pkt = av_packet_alloc(); + if (!pkt) { + av_log(NULL, AV_LOG_ERROR, "Cannot allocate packet\n"); + result = AVERROR(ENOMEM); + goto end; + } + + fr = av_frame_alloc(); + if (!fr) { + av_log(NULL, AV_LOG_ERROR, "Can't allocate frame\n"); + result = AVERROR(ENOMEM); + goto end; + } + + for (;;) { + result = av_read_frame(fmt_ctx, pkt); + if (result) + goto end; + + if (pkt->stream_index != stream_idx) { + av_packet_unref(pkt); + continue; + } + + printf("Stream ID: %d, packet PTS: %s, packet DTS: %s\n", + pkt->stream_index, av_ts2str(pkt->pts), av_ts2str(pkt->dts)); + + if (st->event_flags & AVSTREAM_EVENT_FLAG_METADATA_UPDATED) { + result = av_dict_get_string(st->metadata, &metadata, '=', ':'); + if (result < 0) + goto end; + + printf("Stream ID: %d, new metadata: %s\n", pkt->stream_index, + strlen(metadata) ? metadata : "N/A"); + + st->event_flags &= ~AVSTREAM_EVENT_FLAG_METADATA_UPDATED; + } + + result = avcodec_send_packet(ctx, pkt); + av_packet_unref(pkt); + + if (result < 0) + goto end; + + do { + result = avcodec_receive_frame(ctx, fr); + if (result == AVERROR_EOF) { + result = 0; + goto end; + } + + if (result == AVERROR(EAGAIN)) + break; + + if (result < 0) + goto end; + + result = av_dict_get_string(fr->metadata, &metadata, '=', ':'); + if (result < 0) + goto end; + + printf("Stream ID: %d, frame PTS: %s, metadata: %s\n", + pkt->stream_index, av_ts2str(fr->pts), + strlen(metadata) ? metadata : "N/A"); + } while (1); + } + +end: + av_packet_free(&pkt); + av_frame_free(&fr); + avformat_close_input(&fmt_ctx); + avcodec_free_context(&ctx); + return result; +} + +int main(int argc, char **argv) { + if (argc < 2) { + av_log(NULL, AV_LOG_ERROR, "Incorrect input\n"); + return 1; + } + + if (dump_stream_meta(argv[1]) != AVERROR_EOF) + return 1; + + return 0; +} diff --git a/tests/fate/ogg-flac.mak b/tests/fate/ogg-flac.mak new file mode 100644 index 0000000000..0d6a015161 --- /dev/null +++ b/tests/fate/ogg-flac.mak @@ -0,0 +1,11 @@ +FATE_OGG_FLAC += fate-ogg-flac-chained-meta +fate-ogg-flac-chained-meta: REF = $(SRC_PATH)/tests/ref/fate/ogg-flac-chained-meta.txt +fate-ogg-flac-chained-meta: CMD = run $(APITESTSDIR)/api-dump-stream-meta-test$(EXESUF) $(TARGET_SAMPLES)/ogg-flac/chained-meta.ogg + +FATE_OGG_FLAC-$(call DEMDEC, OGG, FLAC) += $(FATE_OGG_FLAC) + +FATE_SAMPLES_DUMP_STREAM_META += $(FATE_OGG_FLAC-yes) + +FATE_EXTERN += $(FATE_OGG_FLAC-yes) + +fate-ogg-flac: $(FATE_OGG_FLAC-yes) diff --git a/tests/fate/ogg-opus.mak b/tests/fate/ogg-opus.mak new file mode 100644 index 0000000000..54b6fbabde --- /dev/null +++ b/tests/fate/ogg-opus.mak @@ -0,0 +1,11 @@ +FATE_OGG_OPUS += fate-ogg-opus-chained-meta +fate-ogg-opus-chained-meta: REF = $(SRC_PATH)/tests/ref/fate/ogg-opus-chained-meta.txt +fate-ogg-opus-chained-meta: CMD = run $(APITESTSDIR)/api-dump-stream-meta-test$(EXESUF) $(TARGET_SAMPLES)/ogg-opus/chained-meta.ogg + +FATE_OGG_OPUS-$(call DEMDEC, OGG, OPUS) += $(FATE_OGG_OPUS) + +FATE_SAMPLES_DUMP_STREAM_META += $(FATE_OGG_OPUS-yes) + +FATE_EXTERN += $(FATE_OGG_OPUS-yes) + +fate-ogg-opus: $(FATE_OGG_OPUS-yes) diff --git a/tests/fate/ogg-vorbis.mak b/tests/fate/ogg-vorbis.mak new file mode 100644 index 0000000000..74805d591e --- /dev/null +++ b/tests/fate/ogg-vorbis.mak @@ -0,0 +1,11 @@ +FATE_OGG_VORBIS += fate-ogg-vorbis-chained-meta +fate-ogg-vorbis-chained-meta: REF = $(SRC_PATH)/tests/ref/fate/ogg-vorbis-chained-meta.txt +fate-ogg-vorbis-chained-meta: CMD = run $(APITESTSDIR)/api-dump-stream-meta-test$(EXESUF) $(TARGET_SAMPLES)/ogg-vorbis/chained-meta.ogg + +FATE_OGG_VORBIS-$(call DEMDEC, OGG, VORBIS) += $(FATE_OGG_VORBIS) + +FATE_SAMPLES_DUMP_STREAM_META += $(FATE_OGG_VORBIS-yes) + +FATE_EXTERN += $(FATE_OGG_VORBIS-yes) + +fate-ogg-vorbis: $(FATE_OGG_VORBIS-yes) diff --git a/tests/ref/fate/ogg-flac-chained-meta.txt b/tests/ref/fate/ogg-flac-chained-meta.txt new file mode 100644 index 0000000000..ad20ba935f --- /dev/null +++ b/tests/ref/fate/ogg-flac-chained-meta.txt @@ -0,0 +1,12 @@ +Stream ID: 0, codec name: flac, metadata: encoder=Lavc61.19.100 flac:title=First Stream +Stream ID: 0, packet PTS: 0, packet DTS: 0 +Stream ID: 0, new metadata: encoder=Lavc61.19.100 flac:title=First Stream +Stream ID: 0, frame PTS: 0, metadata: N/A +Stream ID: 0, packet PTS: 4608, packet DTS: 4608 +Stream ID: 0, frame PTS: 4608, metadata: N/A +Stream ID: 0, packet PTS: 0, packet DTS: 0 +Stream ID: 0, packet PTS: 0, packet DTS: 0 +Stream ID: 0, packet PTS: 0, packet DTS: 0 +Stream ID: 0, frame PTS: 0, metadata: N/A +Stream ID: 0, packet PTS: 4608, packet DTS: 4608 +Stream ID: 0, frame PTS: 4608, metadata: N/A diff --git a/tests/ref/fate/ogg-opus-chained-meta.txt b/tests/ref/fate/ogg-opus-chained-meta.txt new file mode 100644 index 0000000000..fc84b8b703 --- /dev/null +++ b/tests/ref/fate/ogg-opus-chained-meta.txt @@ -0,0 +1,27 @@ +Stream ID: 0, codec name: opus, metadata: encoder=Lavc61.19.100 libopus:title=First Stream +Stream ID: 0, packet PTS: -312, packet DTS: -312 +Stream ID: 0, new metadata: encoder=Lavc61.19.100 libopus:title=First Stream +Stream ID: 0, frame PTS: -312, metadata: N/A +Stream ID: 0, packet PTS: 648, packet DTS: 648 +Stream ID: 0, frame PTS: 648, metadata: N/A +Stream ID: 0, packet PTS: 1608, packet DTS: 1608 +Stream ID: 0, frame PTS: 1608, metadata: N/A +Stream ID: 0, packet PTS: 2568, packet DTS: 2568 +Stream ID: 0, frame PTS: 2568, metadata: N/A +Stream ID: 0, packet PTS: 3528, packet DTS: 3528 +Stream ID: 0, frame PTS: 3528, metadata: N/A +Stream ID: 0, packet PTS: 4488, packet DTS: 4488 +Stream ID: 0, frame PTS: 4488, metadata: N/A +Stream ID: 0, packet PTS: -312, packet DTS: -312 +Stream ID: 0, new metadata: encoder=Lavc61.19.100 libopus;Lavc61.19.100 libopus:title=First Stream;Second Stream +Stream ID: 0, frame PTS: -312, metadata: N/A +Stream ID: 0, packet PTS: 648, packet DTS: 648 +Stream ID: 0, frame PTS: 648, metadata: N/A +Stream ID: 0, packet PTS: 1608, packet DTS: 1608 +Stream ID: 0, frame PTS: 1608, metadata: N/A +Stream ID: 0, packet PTS: 2568, packet DTS: 2568 +Stream ID: 0, frame PTS: 2568, metadata: N/A +Stream ID: 0, packet PTS: 3528, packet DTS: 3528 +Stream ID: 0, frame PTS: 3528, metadata: N/A +Stream ID: 0, packet PTS: 4488, packet DTS: 4488 +Stream ID: 0, frame PTS: 4488, metadata: N/A diff --git a/tests/ref/fate/ogg-vorbis-chained-meta.txt b/tests/ref/fate/ogg-vorbis-chained-meta.txt new file mode 100644 index 0000000000..b7a97c90e2 --- /dev/null +++ b/tests/ref/fate/ogg-vorbis-chained-meta.txt @@ -0,0 +1,17 @@ +Stream ID: 0, codec name: vorbis, metadata: encoder=Lavc61.19.100 libvorbis:title=First Stream +Stream ID: 0, packet PTS: 0, packet DTS: 0 +Stream ID: 0, new metadata: encoder=Lavc61.19.100 libvorbis:title=First Stream +Stream ID: 0, packet PTS: 128, packet DTS: 128 +Stream ID: 0, frame PTS: 128, metadata: N/A +Stream ID: 0, packet PTS: 704, packet DTS: 704 +Stream ID: 0, frame PTS: 704, metadata: N/A +Stream ID: 0, packet PTS: 0, packet DTS: 0 +Stream ID: 0, packet PTS: 0, packet DTS: 0 +Stream ID: 0, new metadata: encoder=Lavc61.19.100 libvorbis:title=Second Stream +Stream ID: 0, packet PTS: 0, packet DTS: 0 +Stream ID: 0, packet PTS: 0, packet DTS: 0 +Stream ID: 0, frame PTS: 0, metadata: N/A +Stream ID: 0, packet PTS: 128, packet DTS: 128 +Stream ID: 0, frame PTS: 128, metadata: N/A +Stream ID: 0, packet PTS: 704, packet DTS: 704 +Stream ID: 0, frame PTS: 704, metadata: N/A

1 0

avformat/hls: add fmp4 to allowed_extensions
by Michael Niedermayer 29 Apr '25

29 Apr '25

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Sun Apr 6 12:47:34 2025 +0200| [d82016c7302e0ede8626fc3d92f1418c567fbab4] | committer: Michael Niedermayer avformat/hls: add fmp4 to allowed_extensions Fixes: yt-dlp/issues/12700 Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=d82016c7302e0ede8… --- libavformat/hls.c | 1 + 1 file changed, 1 insertion(+) diff --git a/libavformat/hls.c b/libavformat/hls.c index 43b0596df9..f2b98a6d72 100644 --- a/libavformat/hls.c +++ b/libavformat/hls.c @@ -2770,6 +2770,7 @@ static const AVOption hls_options[] = { {.str = "3gp,aac,avi,ac3,eac3,flac,mkv,m3u8,m4a,m4s,m4v,mpg,mov,mp2,mp3,mp4,mpeg,mpegts,ogg,ogv,oga,ts,vob,vtt,wav,webvtt" ",cmfv,cmfa" // Ticket11526 www.nicovideo.jp ",ec3" // part of Ticket11435 (Elisa Viihde (Finnish online recording service)) + ",fmp4" // https://github.com/yt-dlp/yt-dlp/issues/12700 }, INT_MIN, INT_MAX, FLAGS}, {"extension_picky", "Be picky with all extensions matching",

1 0

avformat/hls: Add ec3 to allowed_extensions
by Michael Niedermayer 29 Apr '25

29 Apr '25

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Sun Apr 6 12:43:12 2025 +0200| [68644994fd7cf55613e8b17bcc95b29226285ec5] | committer: Michael Niedermayer avformat/hls: Add ec3 to allowed_extensions Fixes part of Ticket11435 Fixes: Elisa Viihde (Finnish online recording service) Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=68644994fd7cf5561… --- libavformat/hls.c | 1 + 1 file changed, 1 insertion(+) diff --git a/libavformat/hls.c b/libavformat/hls.c index bb09344481..43b0596df9 100644 --- a/libavformat/hls.c +++ b/libavformat/hls.c @@ -2769,6 +2769,7 @@ static const AVOption hls_options[] = { OFFSET(allowed_extensions), AV_OPT_TYPE_STRING, {.str = "3gp,aac,avi,ac3,eac3,flac,mkv,m3u8,m4a,m4s,m4v,mpg,mov,mp2,mp3,mp4,mpeg,mpegts,ogg,ogv,oga,ts,vob,vtt,wav,webvtt" ",cmfv,cmfa" // Ticket11526 www.nicovideo.jp + ",ec3" // part of Ticket11435 (Elisa Viihde (Finnish online recording service)) }, INT_MIN, INT_MAX, FLAGS}, {"extension_picky", "Be picky with all extensions matching",

1 0

avformat/hls: Point user to how to easily contribute a fix for missing extensions
by Michael Niedermayer 29 Apr '25

29 Apr '25

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Sun Apr 6 12:31:19 2025 +0200| [0da6ddfc0135df970f2fca2d110aea78ce73e92c] | committer: Michael Niedermayer avformat/hls: Point user to how to easily contribute a fix for missing extensions Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=0da6ddfc0135df970… --- libavformat/hls.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libavformat/hls.c b/libavformat/hls.c index 679161b390..bb09344481 100644 --- a/libavformat/hls.c +++ b/libavformat/hls.c @@ -745,7 +745,7 @@ static int test_segment(AVFormatContext *s, const AVInputFormat *in_fmt, struct + 2*(ff_match_url_ext(seg->url, c->allowed_extensions) > 0); if (!matchA) { - av_log(s, AV_LOG_ERROR, "URL %s is not in allowed_extensions\n", seg->url); + av_log(s, AV_LOG_ERROR, "URL %s is not in allowed_extensions, consider updating hls.c and submitting a patch to ffmpeg-devel, if this should be added\n", seg->url); return AVERROR_INVALIDDATA; }

1 0

avformat/hls: Add cmfv and cmfa to allowed_extensions
by Michael Niedermayer 29 Apr '25

29 Apr '25

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Sun Apr 6 12:30:04 2025 +0200| [2352145e416c7cbdcf9679ea690c9c1dee4a7936] | committer: Michael Niedermayer avformat/hls: Add cmfv and cmfa to allowed_extensions Fixes: www.nicovideo.jp Fixes: Ticket11526 Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=2352145e416c7cbdc… --- libavformat/hls.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/libavformat/hls.c b/libavformat/hls.c index 6623c80309..679161b390 100644 --- a/libavformat/hls.c +++ b/libavformat/hls.c @@ -2767,7 +2767,9 @@ static const AVOption hls_options[] = { OFFSET(prefer_x_start), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, FLAGS}, {"allowed_extensions", "List of file extensions that hls is allowed to access", OFFSET(allowed_extensions), AV_OPT_TYPE_STRING, - {.str = "3gp,aac,avi,ac3,eac3,flac,mkv,m3u8,m4a,m4s,m4v,mpg,mov,mp2,mp3,mp4,mpeg,mpegts,ogg,ogv,oga,ts,vob,vtt,wav,webvtt"}, + {.str = "3gp,aac,avi,ac3,eac3,flac,mkv,m3u8,m4a,m4s,m4v,mpg,mov,mp2,mp3,mp4,mpeg,mpegts,ogg,ogv,oga,ts,vob,vtt,wav,webvtt" + ",cmfv,cmfa" // Ticket11526 www.nicovideo.jp + }, INT_MIN, INT_MAX, FLAGS}, {"extension_picky", "Be picky with all extensions matching", OFFSET(extension_picky), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, FLAGS},

1 0

avcodec: Fix fate-checkasm-hevc_pel failed on LA.
by Shiyou Yin 29 Apr '25

29 Apr '25

ffmpeg | branch: master | Shiyou Yin <yinshiyou-hf(a)loongson.cn> | Fri Apr 18 09:37:08 2025 +0800| [ab8e160f71f486f516cc06e4a930c4180c3c41df] | committer: Michael Niedermayer avcodec: Fix fate-checkasm-hevc_pel failed on LA. Some loop counters were initialized incorrectly. This patch enhances the handling of loop iterations and residual parts. Reviewed-by: 陈昊 <chenhao(a)loongson.cn> Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=ab8e160f71f486f51… --- libavcodec/loongarch/hevc_mc_bi_lsx.c | 299 ++++++++---------------- libavcodec/loongarch/hevc_mc_uni_lsx.c | 325 +++++++------------------- libavcodec/loongarch/hevcdsp_init_loongarch.c | 1 - libavcodec/loongarch/hevcdsp_lsx.c | 233 ++++++++---------- libavcodec/loongarch/hevcdsp_lsx.h | 1 - 5 files changed, 276 insertions(+), 583 deletions(-) diff --git a/libavcodec/loongarch/hevc_mc_bi_lsx.c b/libavcodec/loongarch/hevc_mc_bi_lsx.c index d7ddd1c246..feb85956ac 100644 --- a/libavcodec/loongarch/hevc_mc_bi_lsx.c +++ b/libavcodec/loongarch/hevc_mc_bi_lsx.c @@ -227,7 +227,7 @@ void hevc_bi_copy_8w_lsx(const uint8_t *src0_ptr, int32_t src_stride, const int16_t *src1_ptr, int32_t src2_stride, uint8_t *dst, int32_t dst_stride, int32_t height) { - int32_t loop_cnt = height >> 3; + uint32_t loop_cnt; int32_t res = (height & 7) >> 1; int32_t src_stride_2x = (src_stride << 1); int32_t dst_stride_2x = (dst_stride << 1); @@ -312,7 +312,8 @@ void hevc_bi_copy_12w_lsx(const uint8_t *src0_ptr, int32_t src_stride, const int16_t *src1_ptr, int32_t src2_stride, uint8_t *dst, int32_t dst_stride, int32_t height) { - uint32_t loop_cnt; + uint32_t loop_cnt = height >> 2; + uint32_t res = (height & 3) >> 1; int32_t src_stride_2x = (src_stride << 1); int32_t dst_stride_2x = (dst_stride << 1); int32_t src_stride_4x = (src_stride << 2); @@ -328,10 +329,9 @@ void hevc_bi_copy_12w_lsx(const uint8_t *src0_ptr, int32_t src_stride, __m128i in0, in1, in2, in3, in4, in5, in6, in7; __m128i dst0, dst1, dst2, dst3, dst4, dst5; - for (loop_cnt = 4; loop_cnt--;) { + for (; loop_cnt--;) { src0 = __lsx_vld(src0_ptr, 0); - DUP2_ARG2(__lsx_vldx, src0_ptr, src_stride, src0_ptr, src_stride_2x, - src1, src2); + DUP2_ARG2(__lsx_vldx, src0_ptr, src_stride, src0_ptr, src_stride_2x, src1, src2); src3 = __lsx_vldx(src0_ptr, src_stride_3x); src0_ptr += src_stride_4x; in0 = __lsx_vld(src1_ptr, 0); @@ -340,8 +340,7 @@ void hevc_bi_copy_12w_lsx(const uint8_t *src0_ptr, int32_t src_stride, in3 = __lsx_vldx(src1_ptr, src2_stride_3x); src1_ptr += src2_stride_2x; in4 = __lsx_vld(_src1, 0); - DUP2_ARG2(__lsx_vldx, _src1, src2_stride_x, _src1, src2_stride_2x, - in5, in6); + DUP2_ARG2(__lsx_vldx, _src1, src2_stride_x, _src1, src2_stride_2x, in5, in6); in7 = __lsx_vldx(_src1, src2_stride_3x); _src1 += src2_stride_2x; @@ -363,6 +362,31 @@ void hevc_bi_copy_12w_lsx(const uint8_t *src0_ptr, int32_t src_stride, __lsx_vstelm_w(out2, dst + dst_stride_3x, 8, 3); dst += dst_stride_4x; } + for (;res--;) { + src0 = __lsx_vld(src0_ptr, 0); + src1 = __lsx_vld(src0_ptr + src_stride, 0); + in0 = __lsx_vld(src1_ptr, 0); + in1 = __lsx_vldx(src1_ptr, src2_stride_x); + dst0 = __lsx_vsllwil_hu_bu(src0, 6); + dst1 = __lsx_vsllwil_hu_bu(src1, 6); + out0 = hevc_bi_rnd_clip(in0, dst0, in1, dst1); + __lsx_vstelm_d(out0, dst, 0, 0); + __lsx_vstelm_d(out0, dst + dst_stride, 0, 1); + + in0 = __lsx_vldrepl_d(_src1, 0); + in1 = __lsx_vldrepl_d(_src1 + src2_stride, 0); + src0 = __lsx_vilvh_w(src1, src0); + in0 = __lsx_vilvl_d(in1, in0); + dst0 = __lsx_vsllwil_hu_bu(src0, 6); + dst0 = __lsx_vsadd_h(dst0, in0); + dst0 = __lsx_vssrarni_bu_h(dst0, dst0, 7); + __lsx_vstelm_w(dst0, dst, 8, 0); + __lsx_vstelm_w(dst0, dst + dst_stride, 8, 1); + src0_ptr += src_stride_2x; + _src1 += src2_stride_x; + src1_ptr += src2_stride_x; + dst += dst_stride_2x; + } } static @@ -370,7 +394,8 @@ void hevc_bi_copy_16w_lsx(const uint8_t *src0_ptr, int32_t src_stride, const int16_t *src1_ptr, int32_t src2_stride, uint8_t *dst, int32_t dst_stride, int32_t height) { - uint32_t loop_cnt; + uint32_t loop_cnt = height >> 2; + uint32_t res = (height & 3) >> 1; int32_t src_stride_2x = (src_stride << 1); int32_t dst_stride_2x = (dst_stride << 1); int32_t src_stride_4x = (src_stride << 2); @@ -387,7 +412,7 @@ void hevc_bi_copy_16w_lsx(const uint8_t *src0_ptr, int32_t src_stride, __m128i dst0_r, dst1_r, dst2_r, dst3_r, dst0_l, dst1_l, dst2_l, dst3_l; __m128i zero = {0}; - for (loop_cnt = (height >> 2); loop_cnt--;) { + for (; loop_cnt--;) { src0 = __lsx_vld(src0_ptr, 0); DUP2_ARG2(__lsx_vldx, src0_ptr, src_stride, src0_ptr, src_stride_2x, src1, src2); @@ -420,6 +445,27 @@ void hevc_bi_copy_16w_lsx(const uint8_t *src0_ptr, int32_t src_stride, __lsx_vstx(out3, dst, dst_stride_3x); dst += dst_stride_4x; } + for (;res--;) { + src0 = __lsx_vld(src0_ptr, 0); + src1 = __lsx_vldx(src0_ptr, src_stride); + in0 = __lsx_vld(src1_ptr, 0); + in1 = __lsx_vldx(src1_ptr, src2_stride_x); + in4 = __lsx_vld(_src1, 0); + in5 = __lsx_vldx(_src1, src2_stride_x); + + DUP2_ARG2(__lsx_vsllwil_hu_bu, src0, 6, src1, 6, dst0_r, dst1_r); + DUP2_ARG2(__lsx_vilvh_b, zero, src0, zero, src1, dst0_l, dst1_l); + DUP2_ARG2(__lsx_vslli_h, dst0_l, 6, dst1_l, 6, dst0_l, dst1_l); + out0 = hevc_bi_rnd_clip(in0, dst0_r, in4, dst0_l); + out1 = hevc_bi_rnd_clip(in1, dst1_r, in5, dst1_l); + __lsx_vst(out0, dst, 0); + __lsx_vstx(out1, dst, dst_stride); + + src0_ptr += src_stride_2x; + _src1 += src2_stride_x; + src1_ptr += src2_stride_x; + dst += dst_stride_2x; + } } static @@ -1061,7 +1107,8 @@ static void hevc_hz_4t_24w_lsx(const uint8_t *src0_ptr, int32_t src_stride, { const int16_t *src1_ptr_tmp; uint8_t *dst_tmp; - uint32_t loop_cnt; + uint32_t loop_cnt = height >> 2; + uint32_t res = (height & 3) >> 1; int32_t dst_stride_2x = (dst_stride << 1); int32_t dst_stride_4x = (dst_stride << 2); int32_t dst_stride_3x = dst_stride_2x + dst_stride; @@ -1086,7 +1133,7 @@ static void hevc_hz_4t_24w_lsx(const uint8_t *src0_ptr, int32_t src_stride, dst_tmp = dst + 16; src1_ptr_tmp = src1_ptr + 16; - for (loop_cnt = (height >> 2); loop_cnt--;) { + for (; loop_cnt--;) { DUP2_ARG2(__lsx_vld, src0_ptr, 0, src0_ptr, 16, src0, src1); src0_ptr += src_stride; DUP2_ARG2(__lsx_vld, src0_ptr, 0, src0_ptr, 16, src2, src3); @@ -1155,6 +1202,42 @@ static void hevc_hz_4t_24w_lsx(const uint8_t *src0_ptr, int32_t src_stride, __lsx_vstelm_d(dst1, dst_tmp + dst_stride_3x, 0, 1); dst_tmp += dst_stride_4x; } + for (; res--;) { + DUP2_ARG2(__lsx_vld, src0_ptr, 0, src0_ptr, 16, src0, src1); + src0_ptr += src_stride; + DUP2_ARG2(__lsx_vld, src0_ptr, 0, src0_ptr, 16, src2, src3); + src0_ptr += src_stride; + DUP2_ARG2(__lsx_vld, src1_ptr, 0, src1_ptr, 16, in0, in1); + src1_ptr += src2_stride; + DUP2_ARG2(__lsx_vld, src1_ptr, 0, src1_ptr, 16, in2, in3); + src1_ptr += src2_stride; + + DUP4_ARG3(__lsx_vshuf_b, src0, src0, mask0, src1, src0, mask2, src2, + src2, mask0, src3, src2, mask2, vec0, vec1, vec2, vec3); + DUP4_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec1, filt0, vec2, filt0, + vec3, filt0, dst0, dst1, dst2, dst3); + DUP4_ARG3(__lsx_vshuf_b, src0, src0, mask1, src1, src0, mask3, src2, + src2, mask1, src3, src2, mask3, vec0, vec1, vec2, vec3); + DUP4_ARG3(__lsx_vdp2add_h_bu_b, dst0, vec0, filt1, dst1, vec1, filt1, + dst2, vec2, filt1, dst3, vec3, filt1, dst0, dst1, dst2, dst3); + dst0 = hevc_bi_rnd_clip(in0, dst0, in1, dst1); + dst1 = hevc_bi_rnd_clip(in2, dst2, in3, dst3); + __lsx_vst(dst0, dst, 0); + __lsx_vstx(dst1, dst, dst_stride); + dst += dst_stride_2x; + + in0 = __lsx_vld(src1_ptr_tmp, 0); + in1 = __lsx_vldx(src1_ptr_tmp, src2_stride_x); + src1_ptr_tmp += src2_stride_x; + DUP2_ARG3(__lsx_vshuf_b, src1, src1, mask0, src3, src3, mask0, vec0, vec1); + DUP2_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec1, filt0, dst0, dst1); + DUP2_ARG3(__lsx_vshuf_b, src1, src1, mask1, src3, src3, mask1, vec0, vec1); + DUP2_ARG3(__lsx_vdp2add_h_bu_b, dst0, vec0, filt1, dst1, vec1, filt1, dst0, dst1); + dst0 = hevc_bi_rnd_clip(in0, dst0, in1, dst1); + __lsx_vstelm_d(dst0, dst_tmp, 0, 0); + __lsx_vstelm_d(dst0, dst_tmp + dst_stride, 0, 1); + dst_tmp += dst_stride_2x; + } } static void hevc_hz_4t_32w_lsx(const uint8_t *src0_ptr, int32_t src_stride, @@ -1206,7 +1289,7 @@ static void hevc_vt_4t_12w_lsx(const uint8_t *src0_ptr, int32_t src_stride, uint8_t *dst, int32_t dst_stride, const int8_t *filter, int32_t height) { - int32_t loop_cnt; + uint32_t loop_cnt; int32_t src_stride_2x = (src_stride << 1); int32_t dst_stride_2x = (dst_stride << 1); int32_t dst_stride_4x = (dst_stride << 2); @@ -1295,7 +1378,7 @@ static void hevc_vt_4t_16w_lsx(const uint8_t *src0_ptr, int32_t src_stride, uint8_t *dst, int32_t dst_stride, const int8_t *filter, int32_t height) { - int32_t loop_cnt; + uint32_t loop_cnt = height >> 2; const int32_t src_stride_2x = (src_stride << 1); const int32_t dst_stride_2x = (dst_stride << 1); const int32_t src_stride_3x = src_stride_2x + src_stride; @@ -1316,7 +1399,7 @@ static void hevc_vt_4t_16w_lsx(const uint8_t *src0_ptr, int32_t src_stride, DUP2_ARG2(__lsx_vilvl_b, src1, src0, src2, src1, src10_r, src21_r); DUP2_ARG2(__lsx_vilvh_b, src1, src0, src2, src1, src10_l, src21_l); - for (loop_cnt = (height >> 2); loop_cnt--;) { + for (; loop_cnt--;) { src3 = __lsx_vld(src0_ptr, 0); src4 = __lsx_vldx(src0_ptr, src_stride); src0_ptr += src_stride_2x; @@ -1480,193 +1563,6 @@ static void hevc_vt_4t_32w_lsx(const uint8_t *src0_ptr, int32_t src_stride, dst + 16, dst_stride, filter, height); } -static void hevc_hv_4t_6w_lsx(const uint8_t *src0_ptr, int32_t src_stride, - const int16_t *src1_ptr, int32_t src2_stride, - uint8_t *dst, int32_t dst_stride, - const int8_t *filter_x, const int8_t *filter_y, - int32_t height) -{ - int32_t src_stride_2x = (src_stride << 1); - int32_t dst_stride_2x = (dst_stride << 1); - int32_t src_stride_4x = (src_stride << 2); - int32_t dst_stride_4x = (dst_stride << 2); - int32_t src2_stride_2x = (src2_stride << 1); - int32_t src2_stride_4x = (src2_stride << 2); - int32_t src_stride_3x = src_stride_2x + src_stride; - int32_t dst_stride_3x = dst_stride_2x + dst_stride; - int32_t src2_stride_3x = src2_stride_2x + src2_stride; - __m128i out0, out1; - __m128i src0, src1, src2, src3, src4, src5, src6; - __m128i vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7, mask1; - __m128i filt0, filt1, filt_h0, filt_h1; - __m128i dsth0, dsth1, dsth2, dsth3, dsth4, dsth5; - __m128i dsth6, dsth7, dsth8, dsth9, dsth10; - __m128i dst0_r, dst0_l, dst1_r, dst1_l, dst2_r, dst2_l, dst3_r, dst3_l; - __m128i dst4_r, dst5_r, dst6_r, dst7_r; - __m128i tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp8; - __m128i reg0, reg1, reg2, reg3; - __m128i mask0 = __lsx_vld(ff_hevc_mask_arr, 0); - - src0_ptr -= (src_stride + 1); - DUP2_ARG2(__lsx_vldrepl_h, filter_x, 0, filter_x, 2, filt0, filt1); - - filt_h1 = __lsx_vld(filter_y, 0); - filt_h1 = __lsx_vsllwil_h_b(filt_h1, 0); - DUP2_ARG2(__lsx_vreplvei_w, filt_h1, 0, filt_h1, 1, filt_h0, filt_h1); - - mask1 = __lsx_vaddi_bu(mask0, 2); - - src0 = __lsx_vld(src0_ptr, 0); - DUP2_ARG2(__lsx_vldx, src0_ptr, src_stride, src0_ptr, src_stride_2x, - src1, src2); - src0_ptr += src_stride_3x; - - DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask0, src0, src0, mask1, vec0, vec1); - DUP2_ARG3(__lsx_vshuf_b, src1, src1, mask0, src1, src1, mask1, vec2, vec3); - DUP2_ARG3(__lsx_vshuf_b, src2, src2, mask0, src2, src2, mask1, vec4, vec5); - - DUP2_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec2, filt0, dsth0, dsth1); - dsth2 = __lsx_vdp2_h_bu_b(vec4, filt0); - DUP2_ARG3(__lsx_vdp2add_h_bu_b, dsth0, vec1, filt1, dsth1, vec3, filt1, - dsth0, dsth1); - dsth2 = __lsx_vdp2add_h_bu_b(dsth2, vec5, filt1); - - DUP2_ARG2(__lsx_vilvl_h, dsth1, dsth0, dsth2, dsth1, tmp0, tmp2); - DUP2_ARG2(__lsx_vilvh_h, dsth1, dsth0, dsth2, dsth1, tmp1, tmp3); - - src3 = __lsx_vld(src0_ptr, 0); - DUP2_ARG2(__lsx_vldx, src0_ptr, src_stride, src0_ptr, src_stride_2x, - src4, src5); - src6 = __lsx_vldx(src0_ptr, src_stride_3x); - src0_ptr += src_stride_4x; - DUP2_ARG3(__lsx_vshuf_b, src3, src3, mask0, src3, src3, mask1, vec0, vec1); - DUP2_ARG3(__lsx_vshuf_b, src4, src4, mask0, src4, src4, mask1, vec2, vec3); - DUP2_ARG3(__lsx_vshuf_b, src5, src5, mask0, src5, src5, mask1, vec4, vec5); - DUP2_ARG3(__lsx_vshuf_b, src6, src6, mask0, src6, src6, mask1, vec6, vec7); - - DUP4_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec2, filt0, vec4, filt0, vec6, - filt0, dsth3, dsth4, dsth5, dsth6); - DUP4_ARG3(__lsx_vdp2add_h_bu_b, dsth3, vec1, filt1, dsth4, vec3, filt1, dsth5, - vec5, filt1, dsth6, vec7, filt1, dsth3, dsth4, dsth5, dsth6); - - src3 = __lsx_vld(src0_ptr, 0); - DUP2_ARG2(__lsx_vldx, src0_ptr, src_stride, src0_ptr, src_stride_2x, - src4, src5); - src6 = __lsx_vldx(src0_ptr, src_stride_3x); - - DUP2_ARG3(__lsx_vshuf_b, src3, src3, mask0, src3, src3, mask1, vec0, vec1); - DUP2_ARG3(__lsx_vshuf_b, src4, src4, mask0, src4, src4, mask1, vec2, vec3); - DUP2_ARG3(__lsx_vshuf_b, src5, src5, mask0, src5, src5, mask1, vec4, vec5); - DUP2_ARG3(__lsx_vshuf_b, src6, src6, mask0, src6, src6, mask1, vec6, vec7); - - DUP4_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec2, filt0, vec4, filt0, vec6, - filt0, dsth7, dsth8, dsth9, dsth10); - DUP4_ARG3(__lsx_vdp2add_h_bu_b, dsth7, vec1, filt1, dsth8, vec3, filt1, dsth9, - vec5, filt1, dsth10, vec7, filt1, dsth7, dsth8, dsth9, dsth10); - - DUP2_ARG2(__lsx_vilvl_h, dsth3, dsth2, dsth4, dsth3, tmp4, tmp6); - DUP2_ARG2(__lsx_vilvh_h, dsth3, dsth2, dsth4, dsth3, tmp5, tmp7); - DUP2_ARG2(__lsx_vilvl_h, dsth5, dsth4, dsth6, dsth5, dsth0, dsth2); - DUP2_ARG2(__lsx_vilvh_h, dsth5, dsth4, dsth6, dsth5, dsth1, dsth3); - DUP4_ARG2(__lsx_vdp2_w_h, tmp0, filt_h0, tmp2, filt_h0, tmp4, filt_h0, - tmp6, filt_h0, dst0_r, dst1_r, dst2_r, dst3_r); - DUP4_ARG3(__lsx_vdp2add_w_h, dst0_r, tmp4, filt_h1, dst1_r, tmp6, - filt_h1, dst2_r, dsth0, filt_h1, dst3_r, dsth2, filt_h1, - dst0_r, dst1_r, dst2_r, dst3_r); - DUP2_ARG2(__lsx_vpickev_d, tmp3, tmp1, tmp7, tmp5, tmp0, tmp8); - dst0_l = __lsx_vdp2_w_h(tmp0, filt_h0); - dst0_l = __lsx_vdp2add_w_h(dst0_l, tmp8, filt_h1); - - DUP2_ARG2(__lsx_vilvl_h, dsth7, dsth6, dsth8, dsth7, tmp0, tmp2); - DUP2_ARG2(__lsx_vilvh_h, dsth7, dsth6, dsth8, dsth7, tmp1, tmp3); - DUP2_ARG2(__lsx_vilvl_h, dsth9, dsth8, dsth10, dsth9, tmp4, tmp6); - DUP2_ARG2(__lsx_vilvh_h, dsth9, dsth8, dsth10, dsth9, tmp5, tmp7); - DUP4_ARG2(__lsx_vdp2_w_h, dsth0, filt_h0, dsth2, filt_h0, tmp0, filt_h0, - tmp2, filt_h0, dst4_r, dst5_r, dst6_r, dst7_r); - DUP4_ARG3(__lsx_vdp2add_w_h, dst4_r, tmp0, filt_h1, dst5_r, tmp2, - filt_h1, dst6_r, tmp4, filt_h1, dst7_r, tmp6, filt_h1, - dst4_r, dst5_r, dst6_r, dst7_r); - DUP2_ARG2(__lsx_vpickev_d, dsth3, dsth1, tmp3, tmp1, tmp0, tmp1); - tmp2 = __lsx_vpickev_d(tmp7, tmp5); - - DUP2_ARG2(__lsx_vdp2_w_h, tmp8, filt_h0, tmp0, filt_h0, dst1_l, dst2_l); - dst3_l = __lsx_vdp2_w_h(tmp1, filt_h0); - DUP2_ARG3(__lsx_vdp2add_w_h, dst1_l, tmp0, filt_h1, dst2_l, tmp1, filt_h1, - dst1_l, dst2_l); - dst3_l = __lsx_vdp2add_w_h(dst3_l, tmp2, filt_h1); - - DUP4_ARG2(__lsx_vsrai_d, dst0_r, 6, dst1_r, 6, dst2_r, 6, dst3_r, 6, - dst0_r, dst1_r, dst2_r, dst3_r); - DUP4_ARG2(__lsx_vsrai_d, dst4_r, 6, dst5_r, 6, dst6_r, 6, dst7_r, 6, - dst4_r, dst5_r, dst6_r, dst7_r); - DUP4_ARG2(__lsx_vsrai_d, dst0_l, 6, dst1_l, 6, dst2_l, 6, dst3_l, 6, - dst0_l, dst1_l, dst2_l, dst3_l); - DUP2_ARG2(__lsx_vpickev_h, dst1_r, dst0_r, dst3_r, dst2_r, tmp0, tmp1); - DUP2_ARG2(__lsx_vpickev_h, dst5_r, dst4_r, dst7_r, dst6_r, tmp2, tmp3); - DUP2_ARG2(__lsx_vpickev_h, dst1_l, dst0_l, dst3_l, dst2_l, tmp4, tmp5); - - reg0 = __lsx_vldrepl_d(src1_ptr, 0); - reg1 = __lsx_vldrepl_d(src1_ptr + src2_stride, 0); - dsth0 = __lsx_vilvl_d(reg1, reg0); - reg0 = __lsx_vldrepl_d(src1_ptr + src2_stride_2x, 0); - reg1 = __lsx_vldrepl_d(src1_ptr + src2_stride_3x, 0); - dsth1 = __lsx_vilvl_d(reg1, reg0); - src1_ptr += src2_stride_4x; - reg0 = __lsx_vldrepl_d(src1_ptr, 0); - reg1 = __lsx_vldrepl_d(src1_ptr + src2_stride, 0); - dsth2 = __lsx_vilvl_d(reg1, reg0); - reg0 = __lsx_vldrepl_d(src1_ptr + src2_stride_2x, 0); - reg1 = __lsx_vldrepl_d(src1_ptr + src2_stride_3x, 0); - dsth3 = __lsx_vilvl_d(reg1, reg0); - - DUP4_ARG2(__lsx_vsadd_h, dsth0, tmp0, dsth1, tmp1, dsth2, tmp2, dsth3, - tmp3, tmp0, tmp1, tmp2, tmp3); - DUP4_ARG2(__lsx_vmaxi_h, tmp0, 0, tmp1, 0, tmp2, 0, tmp3, 0, - tmp0, tmp1, tmp2, tmp3); - DUP2_ARG3(__lsx_vssrlrni_bu_h, tmp1, tmp0, 7, tmp3, tmp2, 7, out0, out1); - - __lsx_vstelm_w(out0, dst, 0, 0); - __lsx_vstelm_w(out0, dst + dst_stride, 0, 1); - __lsx_vstelm_w(out0, dst + dst_stride_2x, 0, 2); - __lsx_vstelm_w(out0, dst + dst_stride_3x, 0, 3); - dst += dst_stride_4x; - __lsx_vstelm_w(out1, dst, 0, 0); - __lsx_vstelm_w(out1, dst + dst_stride, 0, 1); - __lsx_vstelm_w(out1, dst + dst_stride_2x, 0, 2); - __lsx_vstelm_w(out1, dst + dst_stride_3x, 0, 3); - dst -= dst_stride_4x; - - src1_ptr -= src2_stride_4x; - - reg0 = __lsx_vldrepl_w(src1_ptr, 8); - reg1 = __lsx_vldrepl_w(src1_ptr + src2_stride, 8); - reg2 = __lsx_vldrepl_w(src1_ptr + src2_stride_2x, 8); - reg3 = __lsx_vldrepl_w(src1_ptr + src2_stride_3x, 8); - DUP2_ARG2(__lsx_vilvl_w, reg1, reg0, reg3, reg2, tmp0, tmp1); - dsth4 = __lsx_vilvl_d(tmp1, tmp0); - src1_ptr += src2_stride_4x; - - reg0 = __lsx_vldrepl_w(src1_ptr, 8); - reg1 = __lsx_vldrepl_w(src1_ptr + src2_stride, 8); - reg2 = __lsx_vldrepl_w(src1_ptr + src2_stride_2x, 8); - reg3 = __lsx_vldrepl_w(src1_ptr + src2_stride_3x, 8); - DUP2_ARG2(__lsx_vilvl_w, reg1, reg0, reg3, reg2, tmp0, tmp1); - dsth5 = __lsx_vilvl_d(tmp1, tmp0); - DUP2_ARG2(__lsx_vsadd_h, dsth4, tmp4, dsth5, tmp5, tmp4, tmp5); - DUP2_ARG2(__lsx_vmaxi_h, tmp4, 0, tmp5, 7, tmp4, tmp5); - out0 = __lsx_vssrlrni_bu_h(tmp5, tmp4, 7); - - __lsx_vstelm_h(out0, dst, 4, 0); - __lsx_vstelm_h(out0, dst + dst_stride, 4, 1); - __lsx_vstelm_h(out0, dst + dst_stride_2x, 4, 2); - __lsx_vstelm_h(out0, dst + dst_stride_3x, 4, 3); - dst += dst_stride_4x; - __lsx_vstelm_h(out0, dst, 4, 4); - __lsx_vstelm_h(out0, dst + dst_stride, 4, 5); - __lsx_vstelm_h(out0, dst + dst_stride_2x, 4, 6); - __lsx_vstelm_h(out0, dst + dst_stride_3x, 4, 7); -} - static av_always_inline void hevc_hv_4t_8x2_lsx(const uint8_t *src0_ptr, int32_t src_stride, const int16_t *src1_ptr, int32_t src2_stride, uint8_t *dst, int32_t dst_stride, @@ -2281,7 +2177,6 @@ BI_MC_HV(qpel, 48, 8); BI_MC_HV(qpel, 64, 8); BI_MC_HV(epel, 8, 4); -BI_MC_HV(epel, 6, 4); BI_MC_HV(epel, 16, 4); BI_MC_HV(epel, 24, 4); BI_MC_HV(epel, 32, 4); diff --git a/libavcodec/loongarch/hevc_mc_uni_lsx.c b/libavcodec/loongarch/hevc_mc_uni_lsx.c index 6bdc27a824..5ec115bab1 100644 --- a/libavcodec/loongarch/hevc_mc_uni_lsx.c +++ b/libavcodec/loongarch/hevc_mc_uni_lsx.c @@ -545,7 +545,7 @@ static void hevc_hv_8t_64w_lsx(const uint8_t *src, int32_t src_stride, uint8_t * } static av_always_inline -void common_vt_4t_24w_lsx(const uint8_t *src, int32_t src_stride, +void common_vt_2t_24w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, int32_t dst_stride, const int8_t *filter, int32_t height) { @@ -554,8 +554,8 @@ void common_vt_4t_24w_lsx(const uint8_t *src, int32_t src_stride, int32_t src_stride_3x = src_stride_2x + src_stride; const uint8_t *_src; - __m128i src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10; - __m128i src11, filt0, filt1; + __m128i src0, src1, src2, src3, src4, src6, src7, src8, src9, src10; + __m128i filt0, filt1; __m128i src10_r, src32_r, src76_r, src98_r, src21_r, src43_r, src87_r; __m128i src109_r, src10_l, src32_l, src21_l, src43_l; __m128i out0_r, out1_r, out2_r, out3_r, out0_l, out1_l; @@ -578,7 +578,7 @@ void common_vt_4t_24w_lsx(const uint8_t *src, int32_t src_stride, _src += src_stride_3x; DUP2_ARG2(__lsx_vilvl_b, src7, src6, src8, src7, src76_r, src87_r); - for (loop_cnt = 8; loop_cnt--;) { + for (loop_cnt = height >> 1; loop_cnt--;) { /* 16 width */ DUP2_ARG2(__lsx_vld, src, 0, _src, 0, src3, src9); DUP2_ARG2(__lsx_vldx, src, src_stride, _src, src_stride, src4, src10); @@ -605,53 +605,27 @@ void common_vt_4t_24w_lsx(const uint8_t *src, int32_t src_stride, /* 16 + 8 width */ DUP4_ARG3(__lsx_vssrarni_bu_h, out0_l, out0_r, 6, out2_r, out2_r, 6, - out3_r, out3_r, 6, out1_l, out1_r, 6, out1, out2, out3, out4); - __lsx_vst(out1, dst, 0); - __lsx_vstelm_d(out2, dst, 16, 0); - dst += dst_stride; - __lsx_vst(out4, dst, 0); - __lsx_vstelm_d(out3, dst, 16, 0); - dst += dst_stride; - - /* 16 width */ - DUP2_ARG2(__lsx_vld, src, 0, _src, 0, src5, src11); - DUP2_ARG2(__lsx_vldx, src, src_stride, _src, src_stride, src2, src8); - DUP2_ARG2(__lsx_vilvl_b, src5, src4, src2, src5, src10_r, src21_r); - DUP2_ARG2(__lsx_vilvh_b, src5, src4, src2, src5, src10_l, src21_l); - - /* 8 width */ - src += src_stride_2x; - _src += src_stride_2x; - DUP2_ARG2(__lsx_vilvl_b, src11, src10, src8, src11, src76_r, src87_r); - - /* 16 width */ - DUP4_ARG2(__lsx_vdp2_h_bu_b, src32_r, filt0, src32_l, filt0, src43_r, - filt0, src43_l, filt0, out0_r, out0_l, out1_r, out1_l); - DUP4_ARG3(__lsx_vdp2add_h_bu_b, out0_r, src10_r, filt1, out0_l, src10_l, - filt1, out1_r, src21_r, filt1, out1_l, src21_l, filt1, - out0_r, out0_l, out1_r, out1_l); - - /* 8 width */ - DUP2_ARG2(__lsx_vdp2_h_bu_b, src98_r, filt0, src109_r, filt0, - out2_r, out3_r); - DUP2_ARG3(__lsx_vdp2add_h_bu_b, out2_r, src76_r, filt1, out3_r, - src87_r, filt1, out2_r, out3_r); - - /* 16 + 8 width */ - DUP4_ARG3(__lsx_vssrarni_bu_h, out0_l, out0_r, 6, out2_r, out2_r, 6, - out1_l, out1_r, 6, out3_r, out3_r, 6, out1, out2, out3, out4); - + out1_l, out1_r, 6, out3_r, out3_r, 6, out1, out2, out3, out4); __lsx_vst(out1, dst, 0); __lsx_vstelm_d(out2, dst, 16, 0); dst += dst_stride; __lsx_vst(out3, dst, 0); __lsx_vstelm_d(out4, dst, 16, 0); dst += dst_stride; + + src10_r = src32_r; + src21_r = src43_r; + src10_l = src32_l; + src21_l = src43_l; + src2 = src4; + src76_r = src98_r; + src87_r = src109_r; + src8 = src10; } } static av_always_inline -void common_vt_4t_32w_lsx(const uint8_t *src, int32_t src_stride, +void common_vt_2t_32w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, int32_t dst_stride, const int8_t *filter, int32_t height) { @@ -805,102 +779,6 @@ void hevc_hv_4t_8x2_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, __lsx_vstelm_d(out, dst + dst_stride, 0, 1); } -static av_always_inline -void hevc_hv_4t_8multx4_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, - int32_t dst_stride, const int8_t *filter_x, - const int8_t *filter_y, int32_t width8mult) -{ - uint32_t cnt; - const int32_t src_stride_2x = (src_stride << 1); - const int32_t dst_stride_2x = (dst_stride << 1); - const int32_t src_stride_4x = (src_stride << 2); - const int32_t src_stride_3x = src_stride_2x + src_stride; - const int32_t dst_stride_3x = dst_stride_2x + dst_stride; - - __m128i out0, out1; - __m128i src0, src1, src2, src3, src4, src5, src6, mask0, mask1; - __m128i vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7; - __m128i filt0, filt1, filt_h0, filt_h1, filter_vec; - __m128i dst0, dst1, dst2, dst3, dst4, dst5, dst6, tmp0, tmp1, tmp2, tmp3; - __m128i dst0_r, dst0_l, dst1_r, dst1_l, dst2_r, dst2_l, dst3_r, dst3_l; - __m128i dst10_r, dst32_r, dst54_r, dst21_r, dst43_r, dst65_r; - __m128i dst10_l, dst32_l, dst54_l, dst21_l, dst43_l, dst65_l; - - src -= (src_stride + 1); - DUP2_ARG2(__lsx_vldrepl_h, filter_x, 0, filter_x, 2, filt0, filt1); - - filter_vec = __lsx_vld(filter_y, 0); - filter_vec = __lsx_vsllwil_h_b(filter_vec, 0); - DUP2_ARG2(__lsx_vreplvei_w, filter_vec, 0, filter_vec, 1, filt_h0, filt_h1); - - mask0 = __lsx_vld(ff_hevc_mask_arr, 0); - mask1 = __lsx_vaddi_bu(mask0, 2); - - for (cnt = width8mult; cnt--;) { - src0 = __lsx_vld(src, 0); - DUP2_ARG2(__lsx_vldx, src, src_stride, src, src_stride_2x, src1, src2); - src3 = __lsx_vldx(src, src_stride_3x); - src += src_stride_4x; - src4 = __lsx_vld(src, 0); - DUP2_ARG2(__lsx_vldx, src, src_stride, src, src_stride_2x, src5, src6); - src += (8 - src_stride_4x); - DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask0, src0, src0, mask1, - vec0, vec1); - DUP2_ARG3(__lsx_vshuf_b, src1, src1, mask0, src1, src1, mask1, - vec2, vec3); - DUP2_ARG3(__lsx_vshuf_b, src2, src2, mask0, src2, src2, mask1, - vec4, vec5); - - DUP2_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec2, filt0, dst0, dst1); - dst2 = __lsx_vdp2_h_bu_b(vec4, filt0); - DUP2_ARG3(__lsx_vdp2add_h_bu_b, dst0, vec1, filt1, dst1, vec3, filt1, - dst0, dst1); - dst2 = __lsx_vdp2add_h_bu_b(dst2, vec5, filt1); - - DUP2_ARG2(__lsx_vilvl_h, dst1, dst0, dst2, dst1, dst10_r, dst21_r); - DUP2_ARG2(__lsx_vilvh_h, dst1, dst0, dst2, dst1, dst10_l, dst21_l); - - DUP2_ARG3(__lsx_vshuf_b, src3, src3, mask0, src3, src3, mask1, - vec0, vec1); - DUP2_ARG3(__lsx_vshuf_b, src4, src4, mask0, src4, src4, mask1, - vec2, vec3); - DUP2_ARG3(__lsx_vshuf_b, src5, src5, mask0, src5, src5, mask1, - vec4, vec5); - DUP2_ARG3(__lsx_vshuf_b, src6, src6, mask0, src6, src6, mask1, - vec6, vec7); - - DUP4_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec2, filt0, vec4, filt0, - vec6, filt0, dst3, dst4, dst5, dst6); - DUP4_ARG3(__lsx_vdp2add_h_bu_b, dst3, vec1, filt1, dst4, vec3, filt1, - dst5, vec5, filt1, dst6, vec7, filt1, dst3, dst4, dst5, dst6); - - DUP4_ARG2(__lsx_vilvl_h, dst3, dst2, dst4, dst3, dst5, dst4, dst6, - dst5, dst32_r, dst43_r, dst54_r, dst65_r); - DUP4_ARG2(__lsx_vilvh_h, dst3, dst2, dst4, dst3, dst5, dst4, dst6, - dst5, dst32_l, dst43_l, dst54_l, dst65_l); - - DUP4_ARG2(__lsx_vdp2_w_h, dst10_r, filt_h0, dst10_l, filt_h0, dst21_r, - filt_h0, dst21_l, filt_h0, dst0_r, dst0_l, dst1_r, dst1_l); - DUP4_ARG2(__lsx_vdp2_w_h, dst32_r, filt_h0, dst32_l, filt_h0, dst43_r, - filt_h0, dst43_l, filt_h0, dst2_r, dst2_l, dst3_r, dst3_l); - DUP4_ARG3(__lsx_vdp2add_w_h, dst0_r, dst32_r, filt_h1, dst0_l, dst32_l, - filt_h1, dst1_r, dst43_r, filt_h1, dst1_l, dst43_l, filt_h1, - dst0_r, dst0_l, dst1_r, dst1_l); - DUP4_ARG3(__lsx_vdp2add_w_h, dst2_r, dst54_r, filt_h1, dst2_l, dst54_l, - filt_h1, dst3_r, dst65_r, filt_h1, dst3_l, dst65_l, filt_h1, - dst2_r, dst2_l, dst3_r, dst3_l); - - DUP4_ARG3(__lsx_vsrani_h_w, dst0_l, dst0_r, 6, dst1_l, dst1_r, 6, - dst2_l, dst2_r, 6, dst3_l, dst3_r, 6, tmp0, tmp1, tmp2, tmp3); - DUP2_ARG3(__lsx_vssrarni_bu_h, tmp1, tmp0, 6, tmp3, tmp2, 6, out0, out1); - __lsx_vstelm_d(out0, dst, 0, 0); - __lsx_vstelm_d(out0, dst + dst_stride, 0, 1); - __lsx_vstelm_d(out1, dst + dst_stride_2x, 0, 0); - __lsx_vstelm_d(out1, dst + dst_stride_3x, 0, 1); - dst += 8; - } -} - static av_always_inline void hevc_hv_4t_8x6_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, int32_t dst_stride, const int8_t *filter_x, @@ -1009,10 +887,9 @@ void hevc_hv_4t_8x6_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, } static av_always_inline -void hevc_hv_4t_8multx4mult_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, - int32_t dst_stride, const int8_t *filter_x, - const int8_t *filter_y, int32_t height, - int32_t width8mult) +void hevc_hv_4t_8mult_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, + int32_t dst_stride, const int8_t *filter_x, + const int8_t *filter_y, int32_t height, int32_t width8mult) { uint32_t loop_cnt, cnt; const uint8_t *src_tmp; @@ -1134,14 +1011,10 @@ void hevc_hv_4t_8w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, { if (2 == height) { hevc_hv_4t_8x2_lsx(src, src_stride, dst, dst_stride, filter_x, filter_y); - } else if (4 == height) { - hevc_hv_4t_8multx4_lsx(src, src_stride, dst, dst_stride, - filter_x, filter_y, 1); } else if (6 == height) { hevc_hv_4t_8x6_lsx(src, src_stride, dst, dst_stride, filter_x, filter_y); } else if (0 == (height & 0x03)) { - hevc_hv_4t_8multx4mult_lsx(src, src_stride, dst, dst_stride, - filter_x, filter_y, height, 1); + hevc_hv_4t_8mult_lsx(src, src_stride, dst, dst_stride, filter_x, filter_y, height, 1); } } @@ -1160,17 +1033,14 @@ void hevc_hv_4t_12w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, const int32_t src_stride_3x = src_stride_2x + src_stride; const int32_t dst_stride_3x = dst_stride_2x + dst_stride; __m128i out0, out1; - __m128i src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10; + __m128i src0, src1, src2, src3, src4, src5, src6; __m128i vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7; - __m128i mask0, mask1, mask2, mask3; + __m128i mask0, mask1; __m128i filt0, filt1, filt_h0, filt_h1, filter_vec, tmp0, tmp1, tmp2, tmp3; __m128i dsth0, dsth1, dsth2, dsth3, dsth4, dsth5, dsth6; - __m128i dst10, dst21, dst22, dst73, dst84, dst95, dst106; - __m128i dst76_r, dst98_r, dst87_r, dst109_r; __m128i dst10_r, dst32_r, dst54_r, dst21_r, dst43_r, dst65_r; __m128i dst10_l, dst32_l, dst54_l, dst21_l, dst43_l, dst65_l; __m128i dst0_r, dst0_l, dst1_r, dst1_l, dst2_r, dst2_l, dst3_r, dst3_l; - __m128i dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7; src -= (src_stride + 1); DUP2_ARG2(__lsx_vldrepl_h, filter_x, 0, filter_x, 2, filt0, filt1); @@ -1186,8 +1056,7 @@ void hevc_hv_4t_12w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, dst_tmp = dst; src0 = __lsx_vld(src_tmp, 0); - DUP2_ARG2(__lsx_vldx, src_tmp, src_stride, src_tmp, src_stride_2x, - src1, src2); + DUP2_ARG2(__lsx_vldx, src_tmp, src_stride, src_tmp, src_stride_2x, src1, src2); src_tmp += src_stride_3x; DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask0, src0, src0, mask1, vec0, vec1); @@ -1196,17 +1065,15 @@ void hevc_hv_4t_12w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, DUP2_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec2, filt0, dsth0, dsth1); dsth2 = __lsx_vdp2_h_bu_b(vec4, filt0); - DUP2_ARG3(__lsx_vdp2add_h_bu_b, dsth0, vec1, filt1, dsth1, vec3, filt1, - dsth0, dsth1); + DUP2_ARG3(__lsx_vdp2add_h_bu_b, dsth0, vec1, filt1, dsth1, vec3, filt1, dsth0, dsth1); dsth2 = __lsx_vdp2add_h_bu_b(dsth2, vec5, filt1); DUP2_ARG2(__lsx_vilvl_h, dsth1, dsth0, dsth2, dsth1, dst10_r, dst21_r); DUP2_ARG2(__lsx_vilvh_h, dsth1, dsth0, dsth2, dsth1, dst10_l, dst21_l); - for (loop_cnt = 4; loop_cnt--;) { + for (loop_cnt = height >> 2; loop_cnt--;) { src3 = __lsx_vld(src_tmp, 0); - DUP2_ARG2(__lsx_vldx, src_tmp, src_stride, src_tmp, src_stride_2x, - src4, src5); + DUP2_ARG2(__lsx_vldx, src_tmp, src_stride, src_tmp, src_stride_2x, src4, src5); src6 = __lsx_vldx(src_tmp, src_stride_3x); src_tmp += src_stride_4x; @@ -1254,83 +1121,73 @@ void hevc_hv_4t_12w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, dsth2 = dsth6; } - src += 8; - dst += 8; + src_tmp = src + 8; + dst_tmp = dst + 8; - mask2 = __lsx_vld(ff_hevc_mask_arr, 16); - mask3 = __lsx_vaddi_bu(mask2, 2); + src0 = __lsx_vld(src_tmp, 0); + DUP2_ARG2(__lsx_vldx, src_tmp, src_stride, src_tmp, src_stride_2x, src1, src2); + src_tmp += src_stride_3x; - src0 = __lsx_vld(src, 0); - DUP2_ARG2(__lsx_vldx, src, src_stride, src, src_stride_2x, src1, src2); - src += src_stride_3x; - DUP2_ARG3(__lsx_vshuf_b, src1, src0, mask2, src1, src0, mask3, vec0, vec1); - DUP2_ARG3(__lsx_vshuf_b, src2, src1, mask2, src2, src1, mask3, vec2, vec3); + DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask0, src0, src0, mask1, vec0, vec1); + DUP2_ARG3(__lsx_vshuf_b, src1, src1, mask0, src1, src1, mask1, vec2, vec3); + DUP2_ARG3(__lsx_vshuf_b, src2, src2, mask0, src2, src2, mask1, vec4, vec5); - DUP2_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec2, filt0, dst10, dst21); - DUP2_ARG3(__lsx_vdp2add_h_bu_b, dst10, vec1, filt1, dst21, vec3, filt1, - dst10, dst21); + DUP2_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec2, filt0, dsth0, dsth1); + dsth2 = __lsx_vdp2_h_bu_b(vec4, filt0); + DUP2_ARG3(__lsx_vdp2add_h_bu_b, dsth0, vec1, filt1, dsth1, vec3, filt1, dsth0, dsth1); + dsth2 = __lsx_vdp2add_h_bu_b(dsth2, vec5, filt1); - dst10_r = __lsx_vilvl_h(dst21, dst10); - dst21_r = __lsx_vilvh_h(dst21, dst10); - dst22 = __lsx_vreplvei_d(dst21, 1); + DUP2_ARG2(__lsx_vilvl_h, dsth1, dsth0, dsth2, dsth1, dst10_r, dst21_r); + DUP2_ARG2(__lsx_vilvh_h, dsth1, dsth0, dsth2, dsth1, dst10_l, dst21_l); - for (loop_cnt = 2; loop_cnt--;) { - src3 = __lsx_vld(src, 0); - DUP2_ARG2(__lsx_vldx, src, src_stride, src, src_stride_2x, src4, src5); - src6 = __lsx_vldx(src, src_stride_3x); - src += src_stride_4x; - src7 = __lsx_vld(src, 0); - DUP2_ARG2(__lsx_vldx, src, src_stride, src, src_stride_2x, src8, src9); - src10 = __lsx_vldx(src, src_stride_3x); - src += src_stride_4x; - DUP4_ARG3(__lsx_vshuf_b, src7, src3, mask2, src7, src3, mask3, src8, - src4, mask2, src8, src4, mask3, vec0, vec1, vec2, vec3); - DUP4_ARG3(__lsx_vshuf_b, src9, src5, mask2, src9, src5, mask3, src10, - src6, mask2, src10, src6, mask3, vec4, vec5, vec6, vec7); + for (loop_cnt = height >> 2; loop_cnt--;) { + src3 = __lsx_vld(src_tmp, 0); + DUP2_ARG2(__lsx_vldx, src_tmp, src_stride, src_tmp, src_stride_2x, src4, src5); + src6 = __lsx_vldx(src_tmp, src_stride_3x); + src_tmp += src_stride_4x; + + DUP4_ARG3(__lsx_vshuf_b, src3, src3, mask0, src3, src3, mask1, src4, + src4, mask0, src4, src4, mask1, vec0, vec1, vec2, vec3); + DUP4_ARG3(__lsx_vshuf_b, src5, src5, mask0, src5, src5, mask1, src6, + src6, mask0, src6, src6, mask1, vec4, vec5, vec6, vec7); DUP4_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec2, filt0, vec4, filt0, - vec6, filt0, dst73, dst84, dst95, dst106); - DUP4_ARG3(__lsx_vdp2add_h_bu_b, dst73, vec1, filt1, dst84, vec3, - filt1, dst95, vec5, filt1, dst106, vec7, filt1, - dst73, dst84, dst95, dst106); - - dst32_r = __lsx_vilvl_h(dst73, dst22); - DUP2_ARG2(__lsx_vilvl_h, dst84, dst73, dst95, dst84, dst43_r, dst54_r); - DUP2_ARG2(__lsx_vilvh_h, dst84, dst73, dst95, dst84, dst87_r, dst98_r); - dst65_r = __lsx_vilvl_h(dst106, dst95); - dst109_r = __lsx_vilvh_h(dst106, dst95); - dst22 = __lsx_vreplvei_d(dst73, 1); - dst76_r = __lsx_vilvl_h(dst22, dst106); - - DUP4_ARG2(__lsx_vdp2_w_h, dst10_r, filt_h0, dst21_r, filt_h0, dst32_r, - filt_h0, dst43_r, filt_h0, dst0, dst1, dst2, dst3); - DUP4_ARG2(__lsx_vdp2_w_h, dst54_r, filt_h0, dst65_r, filt_h0, dst76_r, - filt_h0, dst87_r, filt_h0, dst4, dst5, dst6, dst7); - DUP4_ARG3(__lsx_vdp2add_w_h, dst0, dst32_r, filt_h1, dst1, dst43_r, - filt_h1, dst2, dst54_r, filt_h1, dst3, dst65_r, filt_h1, - dst0, dst1, dst2, dst3); - DUP4_ARG3(__lsx_vdp2add_w_h, dst4, dst76_r, filt_h1, dst5, dst87_r, - filt_h1, dst6, dst98_r, filt_h1, dst7, dst109_r, filt_h1, - dst4, dst5, dst6, dst7); - - DUP4_ARG3(__lsx_vsrani_h_w, dst1, dst0, 6, dst3, dst2, 6, dst5, dst4, - 6, dst7, dst6, 6, tmp0, tmp1, tmp2, tmp3); + vec6, filt0, dsth3, dsth4, dsth5, dsth6); + DUP4_ARG3(__lsx_vdp2add_h_bu_b, dsth3, vec1, filt1, dsth4, + vec3, filt1, dsth5, vec5, filt1, dsth6, vec7, filt1, + dsth3, dsth4, dsth5, dsth6); + + DUP4_ARG2(__lsx_vilvl_h, dsth3, dsth2, dsth4, dsth3, dsth5, dsth4, + dsth6, dsth5, dst32_r, dst43_r, dst54_r, dst65_r); + DUP4_ARG2(__lsx_vilvh_h, dsth3, dsth2, dsth4, dsth3, dsth5, dsth4, + dsth6, dsth5, dst32_l, dst43_l, dst54_l, dst65_l); + + DUP4_ARG2(__lsx_vdp2_w_h, dst10_r, filt_h0, dst10_l, filt_h0, dst21_r, + filt_h0, dst21_l, filt_h0, dst0_r, dst0_l, dst1_r, dst1_l); + DUP4_ARG2(__lsx_vdp2_w_h, dst32_r, filt_h0, dst32_l, filt_h0, dst43_r, + filt_h0, dst43_l, filt_h0, dst2_r, dst2_l, dst3_r, dst3_l); + DUP4_ARG3(__lsx_vdp2add_w_h, dst0_r, dst32_r, filt_h1, dst0_l, dst32_l, + filt_h1, dst1_r, dst43_r, filt_h1, dst1_l, dst43_l, filt_h1, + dst0_r, dst0_l, dst1_r, dst1_l); + DUP4_ARG3(__lsx_vdp2add_w_h, dst2_r, dst54_r, filt_h1, dst2_l, dst54_l, + filt_h1, dst3_r, dst65_r, filt_h1, dst3_l, dst65_l, filt_h1, + dst2_r, dst2_l, dst3_r, dst3_l); + + DUP4_ARG3(__lsx_vsrani_h_w, dst0_l, dst0_r, 6, dst1_l, dst1_r, 6, dst2_l, + dst2_r, 6, dst3_l, dst3_r, 6, tmp0, tmp1, tmp2, tmp3); DUP2_ARG3(__lsx_vssrarni_bu_h, tmp1, tmp0, 6, tmp3, tmp2, 6, out0, out1); - __lsx_vstelm_w(out0, dst, 0, 0); - __lsx_vstelm_w(out0, dst + dst_stride, 0, 1); - __lsx_vstelm_w(out0, dst + dst_stride_2x, 0, 2); - __lsx_vstelm_w(out0, dst + dst_stride_3x, 0, 3); - dst += dst_stride_4x; - __lsx_vstelm_w(out1, dst, 0, 0); - __lsx_vstelm_w(out1, dst + dst_stride, 0, 1); - __lsx_vstelm_w(out1, dst + dst_stride_2x, 0, 2); - __lsx_vstelm_w(out1, dst + dst_stride_3x, 0, 3); - dst += dst_stride_4x; + __lsx_vstelm_w(out0, dst_tmp, 0, 0); + __lsx_vstelm_w(out0, dst_tmp + dst_stride, 0, 2); + __lsx_vstelm_w(out1, dst_tmp + dst_stride_2x, 0, 0); + __lsx_vstelm_w(out1, dst_tmp + dst_stride_3x, 0, 2); + dst_tmp += dst_stride_4x; - dst10_r = dst98_r; - dst21_r = dst109_r; - dst22 = __lsx_vreplvei_d(dst106, 1); + dst10_r = dst54_r; + dst10_l = dst54_l; + dst21_r = dst65_r; + dst21_l = dst65_l; + dsth2 = dsth6; } } @@ -1338,29 +1195,21 @@ static void hevc_hv_4t_16w_lsx(const uint8_t *src, int32_t src_stride, uint8_t * int32_t dst_stride, const int8_t *filter_x, const int8_t *filter_y, int32_t height) { - if (4 == height) { - hevc_hv_4t_8multx4_lsx(src, src_stride, dst, dst_stride, filter_x, - filter_y, 2); - } else { - hevc_hv_4t_8multx4mult_lsx(src, src_stride, dst, dst_stride, - filter_x, filter_y, height, 2); - } + hevc_hv_4t_8mult_lsx(src, src_stride, dst, dst_stride, filter_x, filter_y, height, 2); } static void hevc_hv_4t_24w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, int32_t dst_stride, const int8_t *filter_x, const int8_t *filter_y, int32_t height) { - hevc_hv_4t_8multx4mult_lsx(src, src_stride, dst, dst_stride, - filter_x, filter_y, height, 3); + hevc_hv_4t_8mult_lsx(src, src_stride, dst, dst_stride, filter_x, filter_y, height, 3); } static void hevc_hv_4t_32w_lsx(const uint8_t *src, int32_t src_stride, uint8_t *dst, int32_t dst_stride, const int8_t *filter_x, const int8_t *filter_y, int32_t height) { - hevc_hv_4t_8multx4mult_lsx(src, src_stride, dst, dst_stride, - filter_x, filter_y, height, 4); + hevc_hv_4t_8mult_lsx(src, src_stride, dst, dst_stride, filter_x, filter_y, height, 4); } #define UNI_MC(PEL, DIR, WIDTH, TAP, DIR1, FILT_DIR) \ @@ -1386,8 +1235,8 @@ UNI_MC(qpel, v, 32, 8, vt, my); UNI_MC(qpel, v, 48, 8, vt, my); UNI_MC(qpel, v, 64, 8, vt, my); -UNI_MC(epel, v, 24, 4, vt, my); -UNI_MC(epel, v, 32, 4, vt, my); +UNI_MC(epel, v, 24, 2, vt, my); +UNI_MC(epel, v, 32, 2, vt, my); #undef UNI_MC diff --git a/libavcodec/loongarch/hevcdsp_init_loongarch.c b/libavcodec/loongarch/hevcdsp_init_loongarch.c index 1585bda276..55fe6b15ad 100644 --- a/libavcodec/loongarch/hevcdsp_init_loongarch.c +++ b/libavcodec/loongarch/hevcdsp_init_loongarch.c @@ -139,7 +139,6 @@ void ff_hevc_dsp_init_loongarch(HEVCDSPContext *c, const int bit_depth) c->put_hevc_epel_bi[6][1][0] = ff_hevc_put_hevc_bi_epel_v24_8_lsx; c->put_hevc_epel_bi[7][1][0] = ff_hevc_put_hevc_bi_epel_v32_8_lsx; - c->put_hevc_epel_bi[2][1][1] = ff_hevc_put_hevc_bi_epel_hv6_8_lsx; c->put_hevc_epel_bi[3][1][1] = ff_hevc_put_hevc_bi_epel_hv8_8_lsx; c->put_hevc_epel_bi[5][1][1] = ff_hevc_put_hevc_bi_epel_hv16_8_lsx; c->put_hevc_epel_bi[6][1][1] = ff_hevc_put_hevc_bi_epel_hv24_8_lsx; diff --git a/libavcodec/loongarch/hevcdsp_lsx.c b/libavcodec/loongarch/hevcdsp_lsx.c index 5747925525..993944164d 100644 --- a/libavcodec/loongarch/hevcdsp_lsx.c +++ b/libavcodec/loongarch/hevcdsp_lsx.c @@ -772,56 +772,40 @@ static void hevc_hz_8t_12w_lsx(const uint8_t *src, int32_t src_stride, DUP2_ARG2(__lsx_vaddi_bu, mask4, 2, mask4, 4, mask5, mask6); mask7 = __lsx_vaddi_bu(mask4, 6); - for (loop_cnt = 4; loop_cnt--;) { + for (loop_cnt = height >> 2; loop_cnt--;) { src0 = __lsx_vld(src, 0); DUP2_ARG2(__lsx_vldx, src, src_stride, src, src_stride_2x, src1, src2); src3 = __lsx_vldx(src, src_stride_3x); src4 = __lsx_vld(_src, 0); - DUP2_ARG2(__lsx_vldx, _src, src_stride, _src, src_stride_2x, - src5, src6); + DUP2_ARG2(__lsx_vldx, _src, src_stride, _src, src_stride_2x, src5, src6); src7 = __lsx_vldx(_src, src_stride_3x); src += src_stride_4x; _src += src_stride_4x; - DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask0, src1, src1, mask0, - vec0, vec1); - DUP2_ARG3(__lsx_vshuf_b, src2, src2, mask0, src3, src3, mask0, - vec2, vec3); - DUP2_ARG3(__lsx_vshuf_b, src5, src4, mask4, src7, src6, mask4, - vec4, vec5); + DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask0, src1, src1, mask0, vec0, vec1); + DUP2_ARG3(__lsx_vshuf_b, src2, src2, mask0, src3, src3, mask0, vec2, vec3); + DUP2_ARG3(__lsx_vshuf_b, src5, src4, mask4, src7, src6, mask4, vec4, vec5); DUP4_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec1, filt0, vec2, filt0, vec3, filt0, dst0, dst1, dst2, dst3); DUP2_ARG2(__lsx_vdp2_h_bu_b, vec4, filt0, vec5, filt0, dst4, dst5); - DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask1, src1, src1, mask1, - vec0, vec1); - DUP2_ARG3(__lsx_vshuf_b, src2, src2, mask1, src3, src3, mask1, - vec2, vec3); - DUP2_ARG3(__lsx_vshuf_b, src5, src4, mask5, src7, src6, mask5, - vec4, vec5); + DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask1, src1, src1, mask1, vec0, vec1); + DUP2_ARG3(__lsx_vshuf_b, src2, src2, mask1, src3, src3, mask1, vec2, vec3); + DUP2_ARG3(__lsx_vshuf_b, src5, src4, mask5, src7, src6, mask5, vec4, vec5); DUP4_ARG3(__lsx_vdp2add_h_bu_b, dst0, vec0, filt1, dst1, vec1, filt1, dst2, vec2, filt1, dst3, vec3, filt1, dst0, dst1, dst2, dst3); - DUP2_ARG3(__lsx_vdp2add_h_bu_b, dst4, vec4, filt1, dst5, vec5, filt1, - dst4, dst5); - DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask2, src1, src1, mask2, - vec0, vec1); - DUP2_ARG3(__lsx_vshuf_b, src2, src2, mask2, src3, src3, mask2, - vec2, vec3); - DUP2_ARG3(__lsx_vshuf_b, src5, src4, mask6, src7, src6, mask6, - vec4, vec5); + DUP2_ARG3(__lsx_vdp2add_h_bu_b, dst4, vec4, filt1, dst5, vec5, filt1, dst4, dst5); + DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask2, src1, src1, mask2, vec0, vec1); + DUP2_ARG3(__lsx_vshuf_b, src2, src2, mask2, src3, src3, mask2, vec2, vec3); + DUP2_ARG3(__lsx_vshuf_b, src5, src4, mask6, src7, src6, mask6, vec4, vec5); DUP4_ARG3(__lsx_vdp2add_h_bu_b, dst0, vec0, filt2, dst1, vec1, filt2, dst2, vec2, filt2, dst3, vec3, filt2, dst0, dst1, dst2, dst3); - DUP2_ARG3(__lsx_vdp2add_h_bu_b, dst4, vec4, filt2, dst5, vec5, filt2, - dst4, dst5); - DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask3, src1, src1, mask3, - vec0, vec1); - DUP2_ARG3(__lsx_vshuf_b, src2, src2, mask3, src3, src3, mask3, - vec2, vec3); - DUP2_ARG3(__lsx_vshuf_b, src5, src4, mask7, src7, src6, mask7, - vec4, vec5); + DUP2_ARG3(__lsx_vdp2add_h_bu_b, dst4, vec4, filt2, dst5, vec5, filt2, dst4, dst5); + DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask3, src1, src1, mask3, vec0, vec1); + DUP2_ARG3(__lsx_vshuf_b, src2, src2, mask3, src3, src3, mask3, vec2, vec3); + DUP2_ARG3(__lsx_vshuf_b, src5, src4, mask7, src7, src6, mask7, vec4, vec5); DUP4_ARG3(__lsx_vdp2add_h_bu_b, dst0, vec0, filt3, dst1, vec1, filt3, dst2, vec2, filt3, dst3, vec3, filt3, dst0, dst1, dst2, dst3); - DUP2_ARG3(__lsx_vdp2add_h_bu_b, dst4, vec4, filt3, dst5, vec5, filt3, - dst4, dst5); + DUP2_ARG3(__lsx_vdp2add_h_bu_b, dst4, vec4, filt3, dst5, vec5, filt3, dst4, dst5); __lsx_vst(dst0, dst, 0); __lsx_vstelm_d(dst4, dst, 16, 0); @@ -2972,16 +2956,14 @@ static void hevc_hv_4t_12w_lsx(const uint8_t *src, int32_t src_stride_3x = src_stride_2x + src_stride; int32_t dst_stride_3x = dst_stride_2x + dst_stride_x; - __m128i src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10; + __m128i src0, src1, src2, src3, src4, src5, src6; __m128i vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7; - __m128i mask0, mask1, mask2, mask3; + __m128i mask0, mask1; __m128i filt0, filt1, filt_h0, filt_h1, filter_vec, dst0; - __m128i dst1, dst2, dst3, dst4, dst5, dst6, dst10, dst21, dst22, dst73; - __m128i dst84, dst95, dst106, dst76_r, dst98_r, dst87_r, dst109_r; + __m128i dst1, dst2, dst3, dst4, dst5, dst6; __m128i dst10_r, dst32_r, dst54_r, dst21_r, dst43_r, dst65_r; __m128i dst10_l, dst32_l, dst54_l, dst21_l, dst43_l, dst65_l; __m128i dst0_r, dst0_l, dst1_r, dst1_l, dst2_r, dst2_l, dst3_r, dst3_l; - __m128i tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7; src -= (src_stride + 1); DUP2_ARG2(__lsx_vldrepl_h, filter_x, 0, filter_x, 2, filt0, filt1); @@ -2997,8 +2979,7 @@ static void hevc_hv_4t_12w_lsx(const uint8_t *src, dst_tmp = dst; src0 = __lsx_vld(src_tmp, 0); - DUP2_ARG2(__lsx_vldx, src_tmp, src_stride, src_tmp, src_stride_2x, - src1, src2); + DUP2_ARG2(__lsx_vldx, src_tmp, src_stride, src_tmp, src_stride_2x, src1, src2); src_tmp += src_stride_3x; DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask0, src0, src0, mask1, vec0, vec1); @@ -3007,34 +2988,27 @@ static void hevc_hv_4t_12w_lsx(const uint8_t *src, DUP2_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec2, filt0, dst0, dst1); dst2 = __lsx_vdp2_h_bu_b(vec4, filt0); - DUP2_ARG3(__lsx_vdp2add_h_bu_b, dst0, vec1, filt1, dst1, vec3, filt1, - dst0, dst1); + DUP2_ARG3(__lsx_vdp2add_h_bu_b, dst0, vec1, filt1, dst1, vec3, filt1, dst0, dst1); dst2 = __lsx_vdp2add_h_bu_b(dst2, vec5, filt1); DUP2_ARG2(__lsx_vilvl_h, dst1, dst0, dst2, dst1, dst10_r, dst21_r); DUP2_ARG2(__lsx_vilvh_h, dst1, dst0, dst2, dst1, dst10_l, dst21_l); - for (loop_cnt = 4; loop_cnt--;) { + for (loop_cnt = height >> 2; loop_cnt--;) { src3 = __lsx_vld(src_tmp, 0); - DUP2_ARG2(__lsx_vldx, src_tmp, src_stride, src_tmp, src_stride_2x, - src4, src5); + DUP2_ARG2(__lsx_vldx, src_tmp, src_stride, src_tmp, src_stride_2x, src4, src5); src6 = __lsx_vldx(src_tmp, src_stride_3x); src_tmp += src_stride_4x; - DUP2_ARG3(__lsx_vshuf_b, src3, src3, mask0, src3, src3, mask1, - vec0, vec1); - DUP2_ARG3(__lsx_vshuf_b, src4, src4, mask0, src4, src4, mask1, - vec2, vec3); - DUP2_ARG3(__lsx_vshuf_b, src5, src5, mask0, src5, src5, mask1, - vec4, vec5); - DUP2_ARG3(__lsx_vshuf_b, src6, src6, mask0, src6, src6, mask1, - vec6, vec7); + DUP2_ARG3(__lsx_vshuf_b, src3, src3, mask0, src3, src3, mask1, vec0, vec1); + DUP2_ARG3(__lsx_vshuf_b, src4, src4, mask0, src4, src4, mask1, vec2, vec3); + DUP2_ARG3(__lsx_vshuf_b, src5, src5, mask0, src5, src5, mask1, vec4, vec5); + DUP2_ARG3(__lsx_vshuf_b, src6, src6, mask0, src6, src6, mask1, vec6, vec7); DUP4_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec2, filt0, vec4, filt0, vec6, filt0, dst3, dst4, dst5, dst6); - DUP4_ARG3(__lsx_vdp2add_h_bu_b, dst3, vec1, filt1, dst4, vec3, - filt1, dst5, vec5, filt1, dst6, vec7, filt1, dst3, - dst4, dst5, dst6); + DUP4_ARG3(__lsx_vdp2add_h_bu_b, dst3, vec1, filt1, dst4, vec3, filt1, + dst5, vec5, filt1, dst6, vec7, filt1, dst3, dst4, dst5, dst6); DUP2_ARG2(__lsx_vilvl_h, dst3, dst2, dst4, dst3, dst32_r, dst43_r); DUP2_ARG2(__lsx_vilvh_h, dst3, dst2, dst4, dst3, dst32_l, dst43_l); @@ -3051,12 +3025,8 @@ static void hevc_hv_4t_12w_lsx(const uint8_t *src, DUP4_ARG3(__lsx_vdp2add_w_h, dst2_r, dst54_r, filt_h1, dst2_l, dst54_l, filt_h1, dst3_r, dst65_r, filt_h1, dst3_l, dst65_l, filt_h1, dst2_r, dst2_l, dst3_r, dst3_l); - DUP4_ARG2(__lsx_vsrai_w, dst0_r, 6, dst0_l, 6, dst1_r, 6, dst1_l, 6, - dst0_r, dst0_l, dst1_r, dst1_l); - DUP4_ARG2(__lsx_vsrai_w, dst2_r, 6, dst2_l, 6, dst3_r, 6, dst3_l, 6, - dst2_r, dst2_l, dst3_r, dst3_l); - DUP4_ARG2(__lsx_vpickev_h, dst0_l, dst0_r, dst1_l, dst1_r, dst2_l, dst2_r, - dst3_l, dst3_r, dst0_r, dst1_r, dst2_r, dst3_r); + DUP2_ARG3(__lsx_vsrani_h_w, dst0_l, dst0_r, 6, dst1_l, dst1_r, 6, dst0_r, dst1_r); + DUP2_ARG3(__lsx_vsrani_h_w, dst2_l, dst2_r, 6, dst3_l, dst3_r, 6, dst2_r, dst3_r); __lsx_vst(dst0_r, dst_tmp, 0); __lsx_vstx(dst1_r, dst_tmp, dst_stride_x); __lsx_vstx(dst2_r, dst_tmp, dst_stride_2x); @@ -3070,93 +3040,74 @@ static void hevc_hv_4t_12w_lsx(const uint8_t *src, dst2 = dst6; } - src += 8; - dst += 8; + src_tmp = src + 8; + dst_tmp = dst + 8; - mask2 = __lsx_vld(ff_hevc_mask_arr, 16); - mask3 = __lsx_vaddi_bu(mask2, 2); + src0 = __lsx_vld(src_tmp, 0); + DUP2_ARG2(__lsx_vldx, src_tmp, src_stride, src_tmp, src_stride_2x, src1, src2); + src_tmp += src_stride_3x; - src0 = __lsx_vld(src, 0); - DUP2_ARG2(__lsx_vldx, src, src_stride, src, src_stride_2x, src1, src2); - src += src_stride_3x; - DUP2_ARG3(__lsx_vshuf_b, src1, src0, mask2, src1, src0, mask3, vec0, vec1); - DUP2_ARG3(__lsx_vshuf_b, src2, src1, mask2, src2, src1, mask3, vec2, vec3); - DUP2_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec2, filt0, dst10, dst21); - DUP2_ARG3(__lsx_vdp2add_h_bu_b, dst10, vec1, filt1, dst21, vec3, filt1, - dst10, dst21); - dst10_r = __lsx_vilvl_h(dst21, dst10); - dst21_r = __lsx_vilvh_h(dst21, dst10); - dst22 = __lsx_vreplvei_d(dst21, 1); - - for (loop_cnt = 2; loop_cnt--;) { - src3 = __lsx_vld(src, 0); - DUP2_ARG2(__lsx_vldx, src, src_stride, src, src_stride_2x, src4, src5); - src6 = __lsx_vldx(src, src_stride_3x); - src += src_stride_4x; - src7 = __lsx_vld(src, 0); - DUP2_ARG2(__lsx_vldx, src, src_stride, src, src_stride_2x, src8, src9); - src10 = __lsx_vldx(src, src_stride_3x); - src += src_stride_4x; - DUP2_ARG3(__lsx_vshuf_b, src7, src3, mask2, src7, src3, mask3, - vec0, vec1); - DUP2_ARG3(__lsx_vshuf_b, src8, src4, mask2, src8, src4, mask3, - vec2, vec3); - DUP2_ARG3(__lsx_vshuf_b, src9, src5, mask2, src9, src5, mask3, - vec4, vec5); - DUP2_ARG3(__lsx_vshuf_b, src10, src6, mask2, src10, src6, mask3, - vec6, vec7); + DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask0, src0, src0, mask1, vec0, vec1); + DUP2_ARG3(__lsx_vshuf_b, src1, src1, mask0, src1, src1, mask1, vec2, vec3); + DUP2_ARG3(__lsx_vshuf_b, src2, src2, mask0, src2, src2, mask1, vec4, vec5); + + DUP2_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec2, filt0, dst0, dst1); + dst2 = __lsx_vdp2_h_bu_b(vec4, filt0); + DUP2_ARG3(__lsx_vdp2add_h_bu_b, dst0, vec1, filt1, dst1, vec3, filt1, dst0, dst1); + dst2 = __lsx_vdp2add_h_bu_b(dst2, vec5, filt1); + + DUP2_ARG2(__lsx_vilvl_h, dst1, dst0, dst2, dst1, dst10_r, dst21_r); + DUP2_ARG2(__lsx_vilvh_h, dst1, dst0, dst2, dst1, dst10_l, dst21_l); + + for (loop_cnt = height >> 2; loop_cnt--;) { + src3 = __lsx_vld(src_tmp, 0); + DUP2_ARG2(__lsx_vldx, src_tmp, src_stride, src_tmp, src_stride_2x, src4, src5); + src6 = __lsx_vldx(src_tmp, src_stride_3x); + src_tmp += src_stride_4x; + + DUP2_ARG3(__lsx_vshuf_b, src3, src3, mask0, src3, src3, mask1, vec0, vec1); + DUP2_ARG3(__lsx_vshuf_b, src4, src4, mask0, src4, src4, mask1, vec2, vec3); + DUP2_ARG3(__lsx_vshuf_b, src5, src5, mask0, src5, src5, mask1, vec4, vec5); + DUP2_ARG3(__lsx_vshuf_b, src6, src6, mask0, src6, src6, mask1, vec6, vec7); DUP4_ARG2(__lsx_vdp2_h_bu_b, vec0, filt0, vec2, filt0, vec4, filt0, - vec6, filt0, dst73, dst84, dst95, dst106); - DUP4_ARG3(__lsx_vdp2add_h_bu_b, dst73, vec1, filt1, dst84, vec3, - filt1, dst95, vec5, filt1, dst106, vec7, filt1, dst73, - dst84, dst95, dst106); - - DUP2_ARG2(__lsx_vilvl_h, dst73, dst22, dst84, dst73, dst32_r, dst43_r); - DUP2_ARG2(__lsx_vilvh_h, dst84, dst73, dst95, dst84, dst87_r, dst98_r); - DUP2_ARG2(__lsx_vilvl_h, dst95, dst84, dst106, dst95, dst54_r, dst65_r); - dst109_r = __lsx_vilvh_h(dst106, dst95); - dst22 = __lsx_vreplvei_d(dst73, 1); - dst76_r = __lsx_vilvl_h(dst22, dst106); + vec6, filt0, dst3, dst4, dst5, dst6); + DUP4_ARG3(__lsx_vdp2add_h_bu_b, dst3, vec1, filt1, dst4, vec3, filt1, + dst5, vec5, filt1, dst6, vec7, filt1, dst3, dst4, dst5, dst6); - DUP4_ARG2(__lsx_vdp2_w_h, dst10_r, filt_h0, dst21_r, filt_h0, dst32_r, - filt_h0, dst43_r, filt_h0, tmp0, tmp1, tmp2, tmp3); - DUP4_ARG2(__lsx_vdp2_w_h, dst54_r, filt_h0, dst65_r, filt_h0, dst76_r, - filt_h0, dst87_r, filt_h0, tmp4, tmp5, tmp6, tmp7); - DUP4_ARG3(__lsx_vdp2add_w_h, tmp0, dst32_r, filt_h1, tmp1, dst43_r, - filt_h1, tmp2, dst54_r, filt_h1, tmp3, dst65_r, filt_h1, - tmp0, tmp1, tmp2, tmp3); - DUP4_ARG3(__lsx_vdp2add_w_h, tmp4, dst76_r, filt_h1, tmp5, dst87_r, - filt_h1, tmp6, dst98_r, filt_h1, tmp7, dst109_r, filt_h1, - tmp4, tmp5, tmp6, tmp7); - DUP4_ARG2(__lsx_vsrai_w, tmp0, 6, tmp1, 6, tmp2, 6, tmp3, 6, - tmp0, tmp1, tmp2, tmp3); - DUP4_ARG2(__lsx_vsrai_w, tmp4, 6, tmp5, 6, tmp6, 6, tmp7, 6, - tmp4, tmp5, tmp6, tmp7); - DUP4_ARG2(__lsx_vpickev_h, tmp1, tmp0, tmp3, tmp2, tmp5, tmp4, - tmp7, tmp6, tmp0, tmp1, tmp2, tmp3); + DUP2_ARG2(__lsx_vilvl_h, dst3, dst2, dst4, dst3, dst32_r, dst43_r); + DUP2_ARG2(__lsx_vilvh_h, dst3, dst2, dst4, dst3, dst32_l, dst43_l); + DUP2_ARG2(__lsx_vilvl_h, dst5, dst4, dst6, dst5, dst54_r, dst65_r); + DUP2_ARG2(__lsx_vilvh_h, dst5, dst4, dst6, dst5, dst54_l, dst65_l); - __lsx_vstelm_d(tmp0, dst, 0, 0); - dst += dst_stride; - __lsx_vstelm_d(tmp0, dst, 0, 1); - dst += dst_stride; - __lsx_vstelm_d(tmp1, dst, 0, 0); - dst += dst_stride; - __lsx_vstelm_d(tmp1, dst, 0, 1); - dst += dst_stride; - __lsx_vstelm_d(tmp2, dst, 0, 0); - dst += dst_stride; - __lsx_vstelm_d(tmp2, dst, 0, 1); - dst += dst_stride; - __lsx_vstelm_d(tmp3, dst, 0, 0); - dst += dst_stride; - __lsx_vstelm_d(tmp3, dst, 0, 1); - dst += dst_stride; + DUP4_ARG2(__lsx_vdp2_w_h, dst10_r, filt_h0, dst10_l, filt_h0, dst21_r, + filt_h0, dst21_l, filt_h0, dst0_r, dst0_l, dst1_r, dst1_l); + DUP4_ARG2(__lsx_vdp2_w_h, dst32_r, filt_h0, dst32_l, filt_h0, dst43_r, + filt_h0, dst43_l, filt_h0, dst2_r, dst2_l, dst3_r, dst3_l); + DUP4_ARG3(__lsx_vdp2add_w_h, dst0_r, dst32_r, filt_h1, dst0_l, dst32_l, + filt_h1, dst1_r, dst43_r, filt_h1, dst1_l, dst43_l, filt_h1, + dst0_r, dst0_l, dst1_r, dst1_l); + DUP4_ARG3(__lsx_vdp2add_w_h, dst2_r, dst54_r, filt_h1, dst2_l, dst54_l, + filt_h1, dst3_r, dst65_r, filt_h1, dst3_l, dst65_l, filt_h1, + dst2_r, dst2_l, dst3_r, dst3_l); + DUP2_ARG3(__lsx_vsrani_h_w, dst0_l, dst0_r, 6, dst1_l, dst1_r, 6, dst0_r, dst1_r); + DUP2_ARG3(__lsx_vsrani_h_w, dst2_l, dst2_r, 6, dst3_l, dst3_r, 6, dst2_r, dst3_r); + __lsx_vstelm_d(dst0_r, dst_tmp, 0, 0); + dst_tmp += dst_stride; + __lsx_vstelm_d(dst1_r, dst_tmp, 0, 0); + dst_tmp += dst_stride; + __lsx_vstelm_d(dst2_r, dst_tmp, 0, 0); + dst_tmp += dst_stride; + __lsx_vstelm_d(dst3_r, dst_tmp, 0, 0); + dst_tmp += dst_stride; - dst10_r = dst98_r; - dst21_r = dst109_r; - dst22 = __lsx_vreplvei_d(dst106, 1); + dst10_r = dst54_r; + dst10_l = dst54_l; + dst21_r = dst65_r; + dst21_l = dst65_l; + dst2 = dst6; } + } static void hevc_hv_4t_16w_lsx(const uint8_t *src, diff --git a/libavcodec/loongarch/hevcdsp_lsx.h b/libavcodec/loongarch/hevcdsp_lsx.h index cf2a519e94..8562a0d681 100644 --- a/libavcodec/loongarch/hevcdsp_lsx.h +++ b/libavcodec/loongarch/hevcdsp_lsx.h @@ -141,7 +141,6 @@ BI_MC(epel, v, 16); BI_MC(epel, v, 24); BI_MC(epel, v, 32); -BI_MC(epel, hv, 6); BI_MC(epel, hv, 8); BI_MC(epel, hv, 16); BI_MC(epel, hv, 24);

1 0

doc/ffprobe: remove entry for show_format_entry
by Gyan Doshi 29 Apr '25

29 Apr '25

ffmpeg | branch: master | Gyan Doshi <ffmpeg(a)gyani.pro> | Tue Apr 29 14:58:04 2025 +0530| [d9bd8e9eb11015b41fa9b2663da1a6455d88cadb] | committer: Gyan Doshi doc/ffprobe: remove entry for show_format_entry The option was removed in 1dd6363581. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=d9bd8e9eb11015b41… --- doc/ffprobe.texi | 7 ------- 1 file changed, 7 deletions(-) diff --git a/doc/ffprobe.texi b/doc/ffprobe.texi index ceedc8793b..8834df8d35 100644 --- a/doc/ffprobe.texi +++ b/doc/ffprobe.texi @@ -139,13 +139,6 @@ stream. All the container format information is printed within a section with name "FORMAT". -@item -show_format_entry @var{name} -Like @option{-show_format}, but only prints the specified entry of the -container format information, rather than all. This option may be given more -than once, then all specified entries will be shown. - -This option is deprecated, use @code{show_entries} instead. - @item -show_entries @var{section_entries} Set list of entries to show.

1 0

avformat/http: Ensure same protocol in ff_http_do_new_request2
by Zhao Zhili 29 Apr '25

29 Apr '25

ffmpeg | branch: master | Zhao Zhili <zhilizhao(a)tencent.com> | Wed Apr 23 20:59:49 2025 +0800| [9187f40e921d5eab8728ffba8e200cfe66e3c008] | committer: Zhao Zhili avformat/http: Ensure same protocol in ff_http_do_new_request2 This can happen for HLS with AES-128 at the middle of m3u8, so old protocol is https while new protocol is crypt+https. And change the log level from ERROR to INFO when protocol/host/port don't match. User should prepared for this function to fail and retry with new connection. Signed-off-by: Zhao Zhili <zhilizhao(a)tencent.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=9187f40e921d5eab8… --- libavformat/http.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/libavformat/http.c b/libavformat/http.c index 65ea5d993c..f7b2a8a029 100644 --- a/libavformat/http.c +++ b/libavformat/http.c @@ -498,8 +498,13 @@ int ff_http_do_new_request2(URLContext *h, const char *uri, AVDictionary **opts) av_url_split(proto2, sizeof(proto2), NULL, 0, hostname2, sizeof(hostname2), &port2, NULL, 0, uri); + if (strcmp(proto1, proto2) != 0) { + av_log(h, AV_LOG_INFO, "Cannot reuse HTTP connection for different protocol %s vs %s\n", + proto1, proto2); + return AVERROR(EINVAL); + } if (port1 != port2 || strncmp(hostname1, hostname2, sizeof(hostname2)) != 0) { - av_log(h, AV_LOG_ERROR, "Cannot reuse HTTP connection for different host: %s:%d != %s:%d\n", + av_log(h, AV_LOG_INFO, "Cannot reuse HTTP connection for different host: %s:%d != %s:%d\n", hostname1, port1, hostname2, port2 );

1 0

avformat/hls: Fix get key file error
by Zhao Zhili 29 Apr '25

29 Apr '25

ffmpeg | branch: master | Zhao Zhili <zhilizhao(a)tencent.com> | Wed Apr 23 20:59:41 2025 +0800| [bb0c4649fb63591a2c9f5e3962815c2db8de50fc] | committer: Zhao Zhili avformat/hls: Fix get key file error The offset and end_offset options are meant for segment, not for key. Signed-off-by: Zhao Zhili <zhilizhao(a)tencent.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=bb0c4649fb63591a2… --- libavformat/hls.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libavformat/hls.c b/libavformat/hls.c index c02add99c0..6623c80309 100644 --- a/libavformat/hls.c +++ b/libavformat/hls.c @@ -1353,7 +1353,7 @@ static int open_input(HLSContext *c, struct playlist *pls, struct segment *seg, if (seg->key_type == KEY_AES_128 || seg->key_type == KEY_SAMPLE_AES) { if (strcmp(seg->key, pls->key_url)) { AVIOContext *pb = NULL; - if (open_url(pls->parent, &pb, seg->key, &c->avio_opts, opts, NULL) == 0) { + if (open_url(pls->parent, &pb, seg->key, &c->avio_opts, NULL, NULL) == 0) { ret = avio_read(pb, pls->key, sizeof(pls->key)); if (ret != sizeof(pls->key)) { av_log(pls->parent, AV_LOG_ERROR, "Unable to read key file %s\n",

1 0