April 2024 - ffmpeg-cvslog - lists.ffmpeg.org

avformat/mov: free the infe allocated item data on failure
by James Almer 30 Apr '24

30 Apr '24

ffmpeg | branch: master | James Almer <jamrial(a)gmail.com> | Fri Apr 26 21:26:01 2024 -0300| [e09164940e4ff0b0ee9228f9d27385211160c6da] | committer: James Almer avformat/mov: free the infe allocated item data on failure Fixes: memleak Fixes: 68212/clusterfuzz-testcase-minimized-ffmpeg_dem_MOV_fuzzer-4963488540721152 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Tested-by: Michael Niedermayer <michael(a)niedermayer.cc> Signed-off-by: James Almer <jamrial(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=e09164940e4ff0b0e… --- libavformat/mov.c | 31 +++++++++++++++++++++++++------ 1 file changed, 25 insertions(+), 6 deletions(-) diff --git a/libavformat/mov.c b/libavformat/mov.c index f903be216d..e8da6c2d65 100644 --- a/libavformat/mov.c +++ b/libavformat/mov.c @@ -84,6 +84,7 @@ typedef struct MOVParseTableEntry { static int mov_read_default(MOVContext *c, AVIOContext *pb, MOVAtom atom); static int mov_read_mfra(MOVContext *c, AVIOContext *f); +static void mov_free_stream_context(AVFormatContext *s, AVStream *st); static int64_t add_ctts_entry(MOVCtts** ctts_data, unsigned int* ctts_count, unsigned int* allocated_size, int count, int duration); @@ -8160,6 +8161,7 @@ static int mov_read_infe(MOVContext *c, AVIOContext *pb, MOVAtom atom, int idx) if (version < 2) { avpriv_report_missing_feature(c->fc, "infe version < 2"); + avio_skip(pb, size); return 1; } @@ -8203,7 +8205,7 @@ static int mov_read_iinf(MOVContext *c, AVIOContext *pb, MOVAtom atom) { HEIFItem *heif_item; int entry_count; - int version, ret; + int version, got_stream = 0, ret, i; if (c->found_iinf) { av_log(c->fc, AV_LOG_WARNING, "Duplicate iinf box found\n"); @@ -8223,20 +8225,33 @@ static int mov_read_iinf(MOVContext *c, AVIOContext *pb, MOVAtom atom) sizeof(*c->heif_item) * (entry_count - c->nb_heif_item)); c->nb_heif_item = FFMAX(c->nb_heif_item, entry_count); - for (int i = 0; i < entry_count; i++) { + for (i = 0; i < entry_count; i++) { MOVAtom infe; infe.size = avio_rb32(pb) - 8; infe.type = avio_rl32(pb); ret = mov_read_infe(c, pb, infe, i); if (ret < 0) - return ret; - if (ret) - return 0; + goto fail; + if (!ret) + got_stream = 1; } - c->found_iinf = 1; + c->found_iinf = got_stream; return 0; +fail: + for (; i >= 0; i--) { + HEIFItem *item = &c->heif_item[i]; + + av_freep(&item->name); + if (!item->st) + continue; + + mov_free_stream_context(c->fc, item->st); + ff_remove_stream(c->fc, item->st); + item->st = NULL; + } + return ret; } static int mov_read_iref_dimg(MOVContext *c, AVIOContext *pb, int version) @@ -9583,6 +9598,10 @@ static int mov_read_header(AVFormatContext *s) return err; } } + // prevent iloc and iinf boxes from being parsed while reading packets. + // this is needed because an iinf box may have been parsed but ignored + // for having old infe boxes which create no streams. + mov->found_iloc = mov->found_iinf = 1; if (pb->seekable & AVIO_SEEKABLE_NORMAL) { if (mov->nb_chapter_tracks > 0 && !mov->ignore_chapters)

1 0

checkasm: vc1dsp: Align buffers sufficiently for the mspel tests
by Martin Storsjö 30 Apr '24

30 Apr '24

ffmpeg | branch: master | Martin Storsjö <martin(a)martin.st> | Tue Apr 30 10:17:50 2024 +0300| [d11be191fa41c652bf6803a7d421a08e6f30d06b] | committer: Martin Storsjö checkasm: vc1dsp: Align buffers sufficiently for the mspel tests This fixes crashes in the mspel tests on x86. Signed-off-by: Martin Storsjö <martin(a)martin.st> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=d11be191fa41c652b… --- tests/checkasm/vc1dsp.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/tests/checkasm/vc1dsp.c b/tests/checkasm/vc1dsp.c index 407d9e5fe8..f18f0f8251 100644 --- a/tests/checkasm/vc1dsp.c +++ b/tests/checkasm/vc1dsp.c @@ -441,10 +441,10 @@ static void check_unescape(void) static void check_mspel_pixels(void) { - LOCAL_ALIGNED_8(uint8_t, src0, [32 * 32]); - LOCAL_ALIGNED_8(uint8_t, src1, [32 * 32]); - LOCAL_ALIGNED_8(uint8_t, dst0, [32 * 32]); - LOCAL_ALIGNED_8(uint8_t, dst1, [32 * 32]); + LOCAL_ALIGNED_16(uint8_t, src0, [32 * 32]); + LOCAL_ALIGNED_16(uint8_t, src1, [32 * 32]); + LOCAL_ALIGNED_16(uint8_t, dst0, [32 * 32]); + LOCAL_ALIGNED_16(uint8_t, dst1, [32 * 32]); VC1DSPContext h;

1 0

avcodec/dovi_rpuenc: avoid division by zero
by Niklas Haas 30 Apr '24

30 Apr '24

ffmpeg | branch: master | Niklas Haas <git(a)haasn.dev> | Tue Apr 30 21:32:04 2024 +0200| [1d89de03d6a9273b5d552a02d50f8f23b6c84f9d] | committer: Niklas Haas avcodec/dovi_rpuenc: avoid division by zero Fixes: https://trac.ffmpeg.org/ticket/10985 > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=1d89de03d6a9273b5… --- libavcodec/dovi_rpuenc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libavcodec/dovi_rpuenc.c b/libavcodec/dovi_rpuenc.c index 8b99bf12d9..3feaa04b9e 100644 --- a/libavcodec/dovi_rpuenc.c +++ b/libavcodec/dovi_rpuenc.c @@ -245,7 +245,7 @@ static inline void put_se_coef(PutBitContext *pb, const AVDOVIRpuDataHeader *hdr static int av_q2den(AVRational q, int den) { - if (q.den == den) + if (!q.den || q.den == den) return q.num; q = av_mul_q(q, av_make_q(den, 1)); return (q.num + (q.den >> 1)) / q.den;

1 0

avformat/mov: support SpatialAudioBox ambisonic layouts with non-diegetic channels
by James Almer 30 Apr '24

30 Apr '24

ffmpeg | branch: master | James Almer <jamrial(a)gmail.com> | Fri Apr 26 11:50:09 2024 -0300| [b7debef3cc84fc4cd68a39d37c8f347c8966dc69] | committer: James Almer avformat/mov: support SpatialAudioBox ambisonic layouts with non-diegetic channels Signed-off-by: James Almer <jamrial(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=b7debef3cc84fc4cd… --- libavformat/mov.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-) diff --git a/libavformat/mov.c b/libavformat/mov.c index d99c971999..f903be216d 100644 --- a/libavformat/mov.c +++ b/libavformat/mov.c @@ -7928,6 +7928,7 @@ static int mov_read_SA3D(MOVContext *c, AVIOContext *pb, MOVAtom atom) AVChannelLayout ch_layout = { 0 }; int ret, i, version, type; int ambisonic_order, channel_order, normalization, channel_count; + int ambi_channels, non_diegetic_channels; if (c->fc->nb_streams < 1) return 0; @@ -7946,11 +7947,12 @@ static int mov_read_SA3D(MOVContext *c, AVIOContext *pb, MOVAtom atom) } type = avio_r8(pb); - if (type) { + if (type & 0x7f) { av_log(c->fc, AV_LOG_WARNING, - "Unsupported ambisonic type %d\n", type); + "Unsupported ambisonic type %d\n", type & 0x7f); return 0; } + non_diegetic_channels = (type >> 7) * 2; // head_locked_stereo ambisonic_order = avio_rb32(pb); @@ -7970,12 +7972,14 @@ static int mov_read_SA3D(MOVContext *c, AVIOContext *pb, MOVAtom atom) channel_count = avio_rb32(pb); if (ambisonic_order < 0 || ambisonic_order > 31 || - channel_count != (ambisonic_order + 1LL) * (ambisonic_order + 1LL)) { + channel_count != ((ambisonic_order + 1LL) * (ambisonic_order + 1LL) + + non_diegetic_channels)) { av_log(c->fc, AV_LOG_ERROR, "Invalid number of channels (%d / %d)\n", channel_count, ambisonic_order); return 0; } + ambi_channels = channel_count - non_diegetic_channels; ret = av_channel_layout_custom_init(&ch_layout, channel_count); if (ret < 0) @@ -7990,7 +7994,10 @@ static int mov_read_SA3D(MOVContext *c, AVIOContext *pb, MOVAtom atom) av_channel_layout_uninit(&ch_layout); return 0; } - ch_layout.u.map[i].id = AV_CHAN_AMBISONIC_BASE + channel; + if (channel >= ambi_channels) + ch_layout.u.map[i].id = channel - ambi_channels; + else + ch_layout.u.map[i].id = AV_CHAN_AMBISONIC_BASE + channel; } ret = av_channel_layout_retype(&ch_layout, 0, AV_CHANNEL_LAYOUT_RETYPE_FLAG_CANONICAL);

1 0

avformat/mov: support SpatialAudioBox ambisonic layouts with arbitrary channel mapping
by James Almer 30 Apr '24

30 Apr '24

ffmpeg | branch: master | James Almer <jamrial(a)gmail.com> | Fri Apr 26 11:37:10 2024 -0300| [37c8d93e56955638983d48e04f08d767eae265e1] | committer: James Almer avformat/mov: support SpatialAudioBox ambisonic layouts with arbitrary channel mapping Signed-off-by: James Almer <jamrial(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=37c8d93e569556389… --- libavformat/mov.c | 29 ++++++++++++++++++++++------- 1 file changed, 22 insertions(+), 7 deletions(-) diff --git a/libavformat/mov.c b/libavformat/mov.c index ede25d3342..d99c971999 100644 --- a/libavformat/mov.c +++ b/libavformat/mov.c @@ -7925,7 +7925,8 @@ cleanup: static int mov_read_SA3D(MOVContext *c, AVIOContext *pb, MOVAtom atom) { AVStream *st; - int i, version, type; + AVChannelLayout ch_layout = { 0 }; + int ret, i, version, type; int ambisonic_order, channel_order, normalization, channel_count; if (c->fc->nb_streams < 1) @@ -7968,24 +7969,38 @@ static int mov_read_SA3D(MOVContext *c, AVIOContext *pb, MOVAtom atom) } channel_count = avio_rb32(pb); - if (ambisonic_order < 0 || channel_count != (ambisonic_order + 1LL) * (ambisonic_order + 1LL)) { + if (ambisonic_order < 0 || ambisonic_order > 31 || + channel_count != (ambisonic_order + 1LL) * (ambisonic_order + 1LL)) { av_log(c->fc, AV_LOG_ERROR, "Invalid number of channels (%d / %d)\n", channel_count, ambisonic_order); return 0; } + ret = av_channel_layout_custom_init(&ch_layout, channel_count); + if (ret < 0) + return 0; + for (i = 0; i < channel_count; i++) { - if (i != avio_rb32(pb)) { - av_log(c->fc, AV_LOG_WARNING, - "Ambisonic channel reordering is not supported\n"); + unsigned channel = avio_rb32(pb); + + if (channel >= channel_count) { + av_log(c->fc, AV_LOG_ERROR, "Invalid channel index (%d / %d)\n", + channel, ambisonic_order); + av_channel_layout_uninit(&ch_layout); return 0; } + ch_layout.u.map[i].id = AV_CHAN_AMBISONIC_BASE + channel; + } + + ret = av_channel_layout_retype(&ch_layout, 0, AV_CHANNEL_LAYOUT_RETYPE_FLAG_CANONICAL); + if (ret < 0) { + av_channel_layout_uninit(&ch_layout); + return 0; } av_channel_layout_uninit(&st->codecpar->ch_layout); - st->codecpar->ch_layout.order = AV_CHANNEL_ORDER_AMBISONIC; - st->codecpar->ch_layout.nb_channels = channel_count; + st->codecpar->ch_layout = ch_layout; return 0; }

1 0

avcodec/riscv/vp9dsp: Fix inclusion guard
by Andreas Rheinhardt 29 Apr '24

29 Apr '24

ffmpeg | branch: master | Andreas Rheinhardt <andreas.rheinhardt(a)outlook.com> | Mon Apr 29 20:32:39 2024 +0200| [08781ebe1aabe99b94e2ee949ca0c95c1443c756] | committer: Andreas Rheinhardt avcodec/riscv/vp9dsp: Fix inclusion guard Fixes fate-source. Reviewed-by: Jan Ekström <jeebjp(a)gmail.com> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt(a)outlook.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=08781ebe1aabe99b9… --- libavcodec/riscv/vp9dsp.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/libavcodec/riscv/vp9dsp.h b/libavcodec/riscv/vp9dsp.h index 0540c7303d..25047ed507 100644 --- a/libavcodec/riscv/vp9dsp.h +++ b/libavcodec/riscv/vp9dsp.h @@ -18,8 +18,8 @@ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ -#ifndef AVCODEC_RISCV_VP9DSP_RISCV_H -#define AVCODEC_RISCV_VP9DSP_RISCV_H +#ifndef AVCODEC_RISCV_VP9DSP_H +#define AVCODEC_RISCV_VP9DSP_H #include <stddef.h> #include <stdint.h> @@ -168,4 +168,4 @@ VP9_COPY_RISCV_RVI_FUNC(4); #undef VP9_BILINEAR_RISCV_RVV_FUNC #undef VP9_COPY_AVG_RISCV_RVV_FUNC -#endif // #ifndef AVCODEC_RISCV_VP9DSP_RISCV_H +#endif // #ifndef AVCODEC_RISCV_VP9DSP_H

1 0

lavc/vp9dsp: R-V V ipred dc
by sunyuechi 29 Apr '24

29 Apr '24

ffmpeg | branch: master | sunyuechi <sunyuechi(a)iscas.ac.cn> | Mon Feb 26 14:42:17 2024 +0800| [c3a96f97f8fda0b16b3c3f83338e98523233b24f] | committer: Rémi Denis-Courmont lavc/vp9dsp: R-V V ipred dc C908: vp9_dc_8x8_8bpp_c: 46.0 vp9_dc_8x8_8bpp_rvv_i64: 41.0 vp9_dc_16x16_8bpp_c: 109.2 vp9_dc_16x16_8bpp_rvv_i32: 72.7 vp9_dc_32x32_8bpp_c: 365.2 vp9_dc_32x32_8bpp_rvv_i32: 165.5 vp9_dc_127_8x8_8bpp_c: 23.0 vp9_dc_127_8x8_8bpp_rvv_i64: 22.0 vp9_dc_127_16x16_8bpp_c: 70.2 vp9_dc_127_16x16_8bpp_rvv_i32: 50.2 vp9_dc_127_32x32_8bpp_c: 295.2 vp9_dc_127_32x32_8bpp_rvv_i32: 136.7 vp9_dc_128_8x8_8bpp_c: 23.0 vp9_dc_128_8x8_8bpp_rvv_i64: 22.0 vp9_dc_128_16x16_8bpp_c: 70.2 vp9_dc_128_16x16_8bpp_rvv_i32: 50.2 vp9_dc_128_32x32_8bpp_c: 295.2 vp9_dc_128_32x32_8bpp_rvv_i32: 136.7 vp9_dc_129_8x8_8bpp_c: 23.0 vp9_dc_129_8x8_8bpp_rvv_i64: 22.0 vp9_dc_129_16x16_8bpp_c: 70.2 vp9_dc_129_16x16_8bpp_rvv_i32: 50.2 vp9_dc_129_32x32_8bpp_c: 295.2 vp9_dc_129_32x32_8bpp_rvv_i32: 136.7 vp9_dc_left_8x8_8bpp_c: 38.0 vp9_dc_left_8x8_8bpp_rvv_i64: 36.0 vp9_dc_left_16x16_8bpp_c: 93.2 vp9_dc_left_16x16_8bpp_rvv_i32: 67.7 vp9_dc_left_32x32_8bpp_c: 333.2 vp9_dc_left_32x32_8bpp_rvv_i32: 158.5 vp9_dc_top_8x8_8bpp_c: 38.7 vp9_dc_top_8x8_8bpp_rvv_i64: 36.0 vp9_dc_top_16x16_8bpp_c: 93.2 vp9_dc_top_16x16_8bpp_rvv_i32: 67.7 vp9_dc_top_32x32_8bpp_c: 333.2 vp9_dc_top_32x32_8bpp_rvv_i32: 156.2 Signed-off-by: Rémi Denis-Courmont <remi(a)remlab.net> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=c3a96f97f8fda0b16… --- libavcodec/riscv/Makefile | 2 + libavcodec/riscv/vp9_intra_rvv.S | 115 ++++++++++++++++++++++++++ libavcodec/riscv/vp9dsp.h | 171 +++++++++++++++++++++++++++++++++++++++ libavcodec/riscv/vp9dsp_init.c | 61 ++++++++++++++ libavcodec/vp9dsp.c | 2 + libavcodec/vp9dsp.h | 1 + 6 files changed, 352 insertions(+) diff --git a/libavcodec/riscv/Makefile b/libavcodec/riscv/Makefile index 6c2ce3001a..69ccd0896d 100644 --- a/libavcodec/riscv/Makefile +++ b/libavcodec/riscv/Makefile @@ -58,5 +58,7 @@ OBJS-$(CONFIG_VC1DSP) += riscv/vc1dsp_init.o RVV-OBJS-$(CONFIG_VC1DSP) += riscv/vc1dsp_rvv.o OBJS-$(CONFIG_VP8DSP) += riscv/vp8dsp_init.o RVV-OBJS-$(CONFIG_VP8DSP) += riscv/vp8dsp_rvv.o +OBJS-$(CONFIG_VP9_DECODER) += riscv/vp9dsp_init.o +RVV-OBJS-$(CONFIG_VP9_DECODER) += riscv/vp9_intra_rvv.o OBJS-$(CONFIG_VORBIS_DECODER) += riscv/vorbisdsp_init.o RVV-OBJS-$(CONFIG_VORBIS_DECODER) += riscv/vorbisdsp_rvv.o diff --git a/libavcodec/riscv/vp9_intra_rvv.S b/libavcodec/riscv/vp9_intra_rvv.S new file mode 100644 index 0000000000..db9774c263 --- /dev/null +++ b/libavcodec/riscv/vp9_intra_rvv.S @@ -0,0 +1,115 @@ +/* + * Copyright (c) 2024 Institue of Software Chinese Academy of Sciences (ISCAS). + * + * 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/riscv/asm.S" + +.macro avgdc size + vwredsumu.vs v16, v8, v16 + vsetivli zero, 1, e16, m1, ta, ma + vmv.x.s t1, v16 + addi t1, t1, 1 << (\size - 1) + srai t1, t1, \size +.endm + +.macro getdc type size +.ifc \type,top + vmv.v.x v16, zero + vle8.v v8, (a3) + avgdc \size +.elseif \type == left + vmv.v.x v16, zero + vle8.v v8, (a2) + avgdc \size +.elseif \type == dc + vmv.v.x v16, zero + vle8.v v8, (a2) + vwredsumu.vs v16, v8, v16 + vle8.v v8, (a3) + avgdc \size +.else + li t1, \type +.endif +.endm + +.macro dc_e32 type size n restore +.ifc \size,32 + li t0, 32 + vsetvli zero, t0, e8, m2, ta, ma +.else + vsetivli zero, 16, e8, m1, ta, ma +.endif + getdc \type \n + +.if \restore == 1 && \size == 32 + vsetvli zero, t0, e8, m2, ta, ma +.elseif \restore == 1 && \size == 16 + vsetivli zero, 16, e8, m1, ta, ma +.endif + vmv.v.x v0, t1 + + .rept \size + vse8.v v0, (a0) + add a0, a0, a1 + .endr + + ret +.endm + +.macro dc_e64 type size n restore + vsetivli zero, 8, e8, mf2, ta, ma + getdc \type \n + + li t0, 64 + vsetvli zero, t0, e8, m4, ta, ma + vmv.v.x v0, t1 + vsetivli zero, 8, e8, mf2, ta, ma + vsse64.v v0, (a0), a1 + + ret +.endm + +.macro func_dc name size type n restore ext +func ff_\()\name\()_\()\size\()x\size\()_rvv, \ext +.ifc \size,8 + dc_e64 \type \size \n \restore +.else + dc_e32 \type \size \n \restore +.endif +endfunc +.endm + +func_dc dc_127 32 127 0 0 zve32x +func_dc dc_127 16 127 0 0 zve32x +func_dc dc_127 8 127 0 0 zve64x +func_dc dc_128 32 128 0 0 zve32x +func_dc dc_128 16 128 0 0 zve32x +func_dc dc_128 8 128 0 0 zve64x +func_dc dc_129 32 129 0 0 zve32x +func_dc dc_129 16 129 0 0 zve32x +func_dc dc_129 8 129 0 0 zve64x +func_dc dc 32 dc 6 1 zve32x +func_dc dc 16 dc 5 1 zve32x +func_dc dc 8 dc 4 0 zve64x +func_dc dc_left 32 left 5 1 zve32x +func_dc dc_left 16 left 4 1 zve32x +func_dc dc_left 8 left 3 0 zve64x +func_dc dc_top 32 top 5 1 zve32x +func_dc dc_top 16 top 4 1 zve32x +func_dc dc_top 8 top 3 0 zve64x diff --git a/libavcodec/riscv/vp9dsp.h b/libavcodec/riscv/vp9dsp.h new file mode 100644 index 0000000000..0540c7303d --- /dev/null +++ b/libavcodec/riscv/vp9dsp.h @@ -0,0 +1,171 @@ +/* + * Copyright (c) 2024 Institue of Software Chinese Academy of Sciences (ISCAS). + * + * 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 + */ + +#ifndef AVCODEC_RISCV_VP9DSP_RISCV_H +#define AVCODEC_RISCV_VP9DSP_RISCV_H + +#include <stddef.h> +#include <stdint.h> + +void ff_dc_32x32_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_top_32x32_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_top_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_top_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_left_32x32_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_left_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_left_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_127_32x32_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_127_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_127_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_128_32x32_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_128_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_128_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_129_32x32_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_129_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_129_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); + +#define VP9_8TAP_RISCV_RVV_FUNC(SIZE, type, type_idx) \ +void ff_put_8tap_##type##_##SIZE##h_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, \ + ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_put_8tap_##type##_##SIZE##v_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, \ + ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_put_8tap_##type##_##SIZE##hv_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, \ + ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_avg_8tap_##type##_##SIZE##h_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, \ + ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_avg_8tap_##type##_##SIZE##v_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, \ + ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_avg_8tap_##type##_##SIZE##hv_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, \ + ptrdiff_t srcstride, \ + int h, int mx, int my); + +#define VP9_BILINEAR_RISCV_RVV_FUNC(SIZE) \ +void ff_put_bilin_##SIZE##h_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_put_bilin_##SIZE##v_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_put_bilin_##SIZE##hv_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_avg_bilin_##SIZE##h_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_avg_bilin_##SIZE##v_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_avg_bilin_##SIZE##hv_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); + +#define VP9_COPY_AVG_RISCV_RVV_FUNC(SIZE) \ +void ff_copy##SIZE##_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_avg##SIZE##_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); + +VP9_8TAP_RISCV_RVV_FUNC(64, regular, FILTER_8TAP_REGULAR); +VP9_8TAP_RISCV_RVV_FUNC(32, regular, FILTER_8TAP_REGULAR); +VP9_8TAP_RISCV_RVV_FUNC(16, regular, FILTER_8TAP_REGULAR); +VP9_8TAP_RISCV_RVV_FUNC(8, regular, FILTER_8TAP_REGULAR); +VP9_8TAP_RISCV_RVV_FUNC(4, regular, FILTER_8TAP_REGULAR); + +VP9_8TAP_RISCV_RVV_FUNC(64, sharp, FILTER_8TAP_SHARP); +VP9_8TAP_RISCV_RVV_FUNC(32, sharp, FILTER_8TAP_SHARP); +VP9_8TAP_RISCV_RVV_FUNC(16, sharp, FILTER_8TAP_SHARP); +VP9_8TAP_RISCV_RVV_FUNC(8, sharp, FILTER_8TAP_SHARP); +VP9_8TAP_RISCV_RVV_FUNC(4, sharp, FILTER_8TAP_SHARP); + +VP9_8TAP_RISCV_RVV_FUNC(64, smooth, FILTER_8TAP_SMOOTH); +VP9_8TAP_RISCV_RVV_FUNC(32, smooth, FILTER_8TAP_SMOOTH); +VP9_8TAP_RISCV_RVV_FUNC(16, smooth, FILTER_8TAP_SMOOTH); +VP9_8TAP_RISCV_RVV_FUNC(8, smooth, FILTER_8TAP_SMOOTH); +VP9_8TAP_RISCV_RVV_FUNC(4, smooth, FILTER_8TAP_SMOOTH); + +VP9_BILINEAR_RISCV_RVV_FUNC(64); +VP9_BILINEAR_RISCV_RVV_FUNC(32); +VP9_BILINEAR_RISCV_RVV_FUNC(16); +VP9_BILINEAR_RISCV_RVV_FUNC(8); +VP9_BILINEAR_RISCV_RVV_FUNC(4); + +VP9_COPY_AVG_RISCV_RVV_FUNC(64); +VP9_COPY_AVG_RISCV_RVV_FUNC(32); +VP9_COPY_AVG_RISCV_RVV_FUNC(16); +VP9_COPY_AVG_RISCV_RVV_FUNC(8); +VP9_COPY_AVG_RISCV_RVV_FUNC(4); + +#define VP9_COPY_RISCV_RVI_FUNC(SIZE) \ +void ff_copy##SIZE##_rvi(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); + +VP9_COPY_RISCV_RVI_FUNC(8); +VP9_COPY_RISCV_RVI_FUNC(4); + +#undef VP9_8TAP_RISCV_RVV_FUNC +#undef VP9_BILINEAR_RISCV_RVV_FUNC +#undef VP9_COPY_AVG_RISCV_RVV_FUNC + +#endif // #ifndef AVCODEC_RISCV_VP9DSP_RISCV_H diff --git a/libavcodec/riscv/vp9dsp_init.c b/libavcodec/riscv/vp9dsp_init.c new file mode 100644 index 0000000000..69ab39004c --- /dev/null +++ b/libavcodec/riscv/vp9dsp_init.c @@ -0,0 +1,61 @@ +/* + * Copyright (c) 2024 Institue of Software Chinese Academy of Sciences (ISCAS). + * + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lervvr 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 + * Lervvr General Public License for more details. + * + * You should have received a copy of the GNU Lervvr 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/attributes.h" +#include "libavutil/cpu.h" +#include "libavutil/riscv/cpu.h" +#include "libavcodec/vp9dsp.h" +#include "vp9dsp.h" + +static av_cold void vp9dsp_intrapred_init_rvv(VP9DSPContext *dsp, int bpp) +{ + #if HAVE_RVV + int flags = av_get_cpu_flags(); + + if (bpp == 8 && flags & AV_CPU_FLAG_RVV_I64 && ff_get_rv_vlenb() >= 16) { + dsp->intra_pred[TX_8X8][DC_PRED] = ff_dc_8x8_rvv; + dsp->intra_pred[TX_8X8][LEFT_DC_PRED] = ff_dc_left_8x8_rvv; + dsp->intra_pred[TX_8X8][DC_127_PRED] = ff_dc_127_8x8_rvv; + dsp->intra_pred[TX_8X8][DC_128_PRED] = ff_dc_128_8x8_rvv; + dsp->intra_pred[TX_8X8][DC_129_PRED] = ff_dc_129_8x8_rvv; + dsp->intra_pred[TX_8X8][TOP_DC_PRED] = ff_dc_top_8x8_rvv; + } + + if (bpp == 8 && flags & AV_CPU_FLAG_RVV_I32 && ff_get_rv_vlenb() >= 16) { + dsp->intra_pred[TX_32X32][DC_PRED] = ff_dc_32x32_rvv; + dsp->intra_pred[TX_16X16][DC_PRED] = ff_dc_16x16_rvv; + dsp->intra_pred[TX_32X32][LEFT_DC_PRED] = ff_dc_left_32x32_rvv; + dsp->intra_pred[TX_16X16][LEFT_DC_PRED] = ff_dc_left_16x16_rvv; + dsp->intra_pred[TX_32X32][DC_127_PRED] = ff_dc_127_32x32_rvv; + dsp->intra_pred[TX_16X16][DC_127_PRED] = ff_dc_127_16x16_rvv; + dsp->intra_pred[TX_32X32][DC_128_PRED] = ff_dc_128_32x32_rvv; + dsp->intra_pred[TX_16X16][DC_128_PRED] = ff_dc_128_16x16_rvv; + dsp->intra_pred[TX_32X32][DC_129_PRED] = ff_dc_129_32x32_rvv; + dsp->intra_pred[TX_16X16][DC_129_PRED] = ff_dc_129_16x16_rvv; + dsp->intra_pred[TX_32X32][TOP_DC_PRED] = ff_dc_top_32x32_rvv; + dsp->intra_pred[TX_16X16][TOP_DC_PRED] = ff_dc_top_16x16_rvv; + } + #endif +} + +av_cold void ff_vp9dsp_init_riscv(VP9DSPContext *dsp, int bpp, int bitexact) +{ + vp9dsp_intrapred_init_rvv(dsp, bpp); +} diff --git a/libavcodec/vp9dsp.c b/libavcodec/vp9dsp.c index d8ddf74d4f..967e6e1e1a 100644 --- a/libavcodec/vp9dsp.c +++ b/libavcodec/vp9dsp.c @@ -100,6 +100,8 @@ av_cold void ff_vp9dsp_init(VP9DSPContext *dsp, int bpp, int bitexact) ff_vp9dsp_init_aarch64(dsp, bpp); #elif ARCH_ARM ff_vp9dsp_init_arm(dsp, bpp); +#elif ARCH_RISCV + ff_vp9dsp_init_riscv(dsp, bpp, bitexact); #elif ARCH_X86 ff_vp9dsp_init_x86(dsp, bpp, bitexact); #elif ARCH_MIPS diff --git a/libavcodec/vp9dsp.h b/libavcodec/vp9dsp.h index be0ac0b181..772848e349 100644 --- a/libavcodec/vp9dsp.h +++ b/libavcodec/vp9dsp.h @@ -131,6 +131,7 @@ void ff_vp9dsp_init_12(VP9DSPContext *dsp); void ff_vp9dsp_init_aarch64(VP9DSPContext *dsp, int bpp); void ff_vp9dsp_init_arm(VP9DSPContext *dsp, int bpp); +void ff_vp9dsp_init_riscv(VP9DSPContext *dsp, int bpp, int bitexact); void ff_vp9dsp_init_x86(VP9DSPContext *dsp, int bpp, int bitexact); void ff_vp9dsp_init_mips(VP9DSPContext *dsp, int bpp); void ff_vp9dsp_init_loongarch(VP9DSPContext *dsp, int bpp);

1 0

checkasm/vc1dsp: add mspel_pixels test
by sunyuechi 29 Apr '24

29 Apr '24

ffmpeg | branch: master | sunyuechi <sunyuechi(a)iscas.ac.cn> | Fri Mar 1 19:45:53 2024 +0800| [dedc2456bf97003323fbfd61f94a6b345062e7d4] | committer: Rémi Denis-Courmont checkasm/vc1dsp: add mspel_pixels test Signed-off-by: Rémi Denis-Courmont <remi(a)remlab.net> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=dedc2456bf9700332… --- tests/checkasm/vc1dsp.c | 37 +++++++++++++++++++++++++++++++++++++ 1 file changed, 37 insertions(+) diff --git a/tests/checkasm/vc1dsp.c b/tests/checkasm/vc1dsp.c index 8ad2b51acf..407d9e5fe8 100644 --- a/tests/checkasm/vc1dsp.c +++ b/tests/checkasm/vc1dsp.c @@ -439,6 +439,40 @@ static void check_unescape(void) } } +static void check_mspel_pixels(void) +{ + LOCAL_ALIGNED_8(uint8_t, src0, [32 * 32]); + LOCAL_ALIGNED_8(uint8_t, src1, [32 * 32]); + LOCAL_ALIGNED_8(uint8_t, dst0, [32 * 32]); + LOCAL_ALIGNED_8(uint8_t, dst1, [32 * 32]); + + VC1DSPContext h; + + const test tests[] = { + VC1DSP_SIZED_TEST(put_vc1_mspel_pixels_tab[0][0], 16, 16) + VC1DSP_SIZED_TEST(put_vc1_mspel_pixels_tab[1][0], 8, 8) + VC1DSP_SIZED_TEST(avg_vc1_mspel_pixels_tab[0][0], 16, 16) + VC1DSP_SIZED_TEST(avg_vc1_mspel_pixels_tab[1][0], 8, 8) + }; + + ff_vc1dsp_init(&h); + + for (size_t t = 0; t < FF_ARRAY_ELEMS(tests); ++t) { + void (*func)(uint8_t *, const uint8_t*, ptrdiff_t, int) = *(void **)((intptr_t) &h + tests[t].offset); + if (check_func(func, "vc1dsp.%s", tests[t].name)) { + declare_func_emms(AV_CPU_FLAG_MMX, void, uint8_t *, const uint8_t*, ptrdiff_t, int); + RANDOMIZE_BUFFER8(dst, 32 * 32); + RANDOMIZE_BUFFER8(src, 32 * 32); + call_ref(dst0, src0, 32, 0); + call_new(dst1, src1, 32, 0); + if (memcmp(dst0, dst1, 32 * 32)) { + fail(); + } + bench_new(dst1, src0, 32, 0); + } + } +} + void checkasm_check_vc1dsp(void) { check_inv_trans_inplace(); @@ -450,4 +484,7 @@ void checkasm_check_vc1dsp(void) check_unescape(); report("unescape_buffer"); + + check_mspel_pixels(); + report("mspel_pixels"); }

1 0

opustab: macro constant as a double
by Lynne 29 Apr '24

29 Apr '24

ffmpeg | branch: master | Lynne <dev(a)lynne.ee> | Mon Apr 29 01:44:46 2024 +0200| [f492095bd3e22ecf5565a08437a0816910bac949] | committer: Lynne opustab: macro constant as a double May increase intermediate precision on some compilers. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=f492095bd3e22ecf5… --- libavcodec/opustab.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libavcodec/opustab.c b/libavcodec/opustab.c index 47624fe651..917375253e 100644 --- a/libavcodec/opustab.c +++ b/libavcodec/opustab.c @@ -1164,7 +1164,7 @@ const uint32_t * const ff_celt_pvq_u_row[15] = { * libopus uses a slighly rounded constant, set to 0.85 exactly, * to simplify its fixed-point version, but it's not significant to impact * compliance. */ -#define CELT_EMPH_COEFF 0.8500061035f +#define CELT_EMPH_COEFF 0.8500061035 DECLARE_ALIGNED(16, const float, ff_opus_deemph_weights)[] = { CELT_EMPH_COEFF,

1 0

avformat/mov: don't read key_size bytes twice in the keys atom
by James Almer 29 Apr '24

29 Apr '24

ffmpeg | branch: release/7.0 | James Almer <jamrial(a)gmail.com> | Mon Apr 1 23:54:53 2024 -0300| [fb8f0ea7b31d78e8a14d79549a76fd71de390907] | committer: James Almer avformat/mov: don't read key_size bytes twice in the keys atom We only support mdta as type, yet we were not skipping other types, but rather reading key_size worth of bytes twice per entry. Signed-off-by: James Almer <jamrial(a)gmail.com> (cherry picked from commit 5a06d3810e41134ee9c2941cc0b371da62b539db) > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=fb8f0ea7b31d78e8a… --- libavformat/mov.c | 1 + 1 file changed, 1 insertion(+) diff --git a/libavformat/mov.c b/libavformat/mov.c index d551a0f8e0..056890c85b 100644 --- a/libavformat/mov.c +++ b/libavformat/mov.c @@ -5048,6 +5048,7 @@ static int mov_read_keys(MOVContext *c, AVIOContext *pb, MOVAtom atom) key_size -= 8; if (type != MKTAG('m','d','t','a')) { avio_skip(pb, key_size); + continue; } c->meta_keys[i] = av_mallocz(key_size + 1); if (!c->meta_keys[i])

1 0