jeans/LibreELEC.tv
jeans/LibreELEC.tv
1
0
Fork 0
mirror of https://github.com/LibreELEC/LibreELEC.tv.git synced 2025-07-24 11:16:51 +00:00
Code Issues Packages Projects Releases Wiki Activity

remove package 'ffmpeg'

Browse Source 
Signed-off-by: Stephan Raue <stephan@openelec.tv>
This commit is contained in:
Stephan Raue 2014-02-03 16:34:33 +01:00
parent 407d1448a8
commit 9dddc7045e
55 changed files with 0 additions and 5383 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

230
packages/multimedia/ffmpeg/package.mk
View File

@ -1,230 +0,0 @@
################################################################################
# This file is part of OpenELEC - http://www.openelec.tv
# Copyright (C) 2009-2014 Stephan Raue (stephan@openelec.tv)
#
# OpenELEC is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 2 of the License, or
# (at your option) any later version.
#
# OpenELEC 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 General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with OpenELEC. If not, see <http://www.gnu.org/licenses/>.
################################################################################
PKG_NAME="ffmpeg"
PKG_VERSION="1.2.5"
PKG_REV="1"
PKG_ARCH="any"
PKG_LICENSE="LGPL"
PKG_SITE="http://ffmpeg.org"
PKG_URL="http://ffmpeg.org/releases/$PKG_NAME-$PKG_VERSION.tar.bz2"
PKG_DEPENDS_TARGET="toolchain yasm:host zlib bzip2 libvorbis gnutls"
PKG_PRIORITY="optional"
PKG_SECTION="multimedia"
PKG_SHORTDESC="FFmpeg is a complete, cross-platform solution to record, convert and stream audio and video."
PKG_LONGDESC="FFmpeg is a complete, cross-platform solution to record, convert and stream audio and video."
PKG_IS_ADDON="no"
PKG_AUTORECONF="no"
if [ "$VAAPI" = yes ]; then
# configure GPU drivers and dependencies:
get_graphicdrivers
PKG_DEPENDS_TARGET="$PKG_DEPENDS_TARGET $LIBVA"
FFMPEG_VAAPI="--enable-vaapi"
else
FFMPEG_VAAPI="--disable-vaapi"
fi
if [ "$VDPAU" = yes ]; then
PKG_DEPENDS_TARGET="$PKG_DEPENDS_TARGET libvdpau"
FFMPEG_VDPAU="--enable-vdpau"
else
FFMPEG_VDPAU="--disable-vdpau"
fi
if [ "$DEBUG" = yes ]; then
FFMPEG_DEBUG="--enable-debug --disable-stripping"
else
FFMPEG_DEBUG="--disable-debug --enable-stripping"
fi
if [ "$OPTIMIZATIONS" = size ]; then
FFMPEG_OPTIM="--disable-small"
else
FFMPEG_OPTIM="--disable-small"
fi
if [ "$CRYSTALHD" = yes ]; then
# disabled, we use XBMC's internal solution
FFMPEG_CRYSTALHD="--disable-crystalhd"
else
FFMPEG_CRYSTALHD="--disable-crystalhd"
fi
case "$TARGET_ARCH" in
arm)
FFMPEG_CPU=""
FFMPEG_TABLES="--enable-hardcoded-tables"
FFMPEG_PIC="--enable-pic"
;;
i?86)
FFMPEG_CPU=""
FFMPEG_TABLES="--disable-hardcoded-tables"
FFMPEG_PIC="--disable-pic"
;;
x86_64)
FFMPEG_CPU=""
FFMPEG_TABLES="--disable-hardcoded-tables"
FFMPEG_PIC="--enable-pic"
;;
esac
case "$TARGET_FPU" in
neon*)
FFMPEG_FPU="--enable-neon"
;;
vfp*)
FFMPEG_FPU=""
;;
*)
FFMPEG_FPU=""
;;
esac
pre_configure_target() {
cd $ROOT/$PKG_BUILD
rm -rf .$TARGET_NAME
export pkg_config="$ROOT/$TOOLCHAIN/bin/pkg-config"
# ffmpeg fails building with LTO support
strip_lto
# ffmpeg fails running with GOLD support
strip_gold
}
configure_target() {
./configure --prefix=/usr \
--cpu=$TARGET_CPU \
--arch=$TARGET_ARCH \
--enable-cross-compile \
--cross-prefix=$TARGET_PREFIX \
--sysroot=$SYSROOT_PREFIX \
--sysinclude="$SYSROOT_PREFIX/usr/include" \
--target-os="linux" \
--nm="$NM" \
--ar="$AR" \
--as="$CC" \
--cc="$CC" \
--ld="$CC" \
--host-cc="$HOST_CC" \
--host-cflags="$HOST_CFLAGS" \
--host-ldflags="$HOST_LDFLAGS" \
--host-libs="-lm" \
--extra-cflags="$CFLAGS" \
--extra-ldflags="$LDFLAGS" \
--extra-libs="" \
--extra-version="" \
--build-suffix="" \
--disable-static \
--enable-shared \
--enable-gpl \
--disable-version3 \
--disable-nonfree \
--enable-logging \
--disable-doc \
$FFMPEG_DEBUG \
$FFMPEG_PIC \
--enable-optimizations \
--disable-extra-warnings \
--disable-ffprobe \
--disable-ffplay \
--disable-ffserver \
--enable-ffmpeg \
--enable-avdevice \
--enable-avcodec \
--enable-avformat \
--enable-swscale \
--enable-postproc \
--enable-avfilter \
--disable-devices \
--enable-pthreads \
--disable-w32threads \
--disable-x11grab \
--enable-network \
--enable-gnutls \
--disable-gray \
--enable-swscale-alpha \
$FFMPEG_OPTIM \
--enable-dct \
--enable-fft \
--enable-mdct \
--enable-rdft \
$FFMPEG_CRYSTALHD \
$FFMPEG_VAAPI \
$FFMPEG_VDPAU \
--disable-dxva2 \
--enable-runtime-cpudetect \
$FFMPEG_TABLES \
--disable-memalign-hack \
--disable-encoders \
--enable-encoder=ac3 \
--enable-encoder=aac \
--disable-decoder=mpeg_xvmc \
--enable-hwaccels \
--disable-muxers \
--enable-muxer=spdif \
--enable-muxer=adts \
--enable-demuxers \
--enable-parsers \
--enable-bsfs \
--enable-protocol=http \
--disable-indevs \
--disable-outdevs \
--enable-filters \
--disable-avisynth \
--enable-bzlib \
--disable-frei0r \
--disable-libopencore-amrnb \
--disable-libopencore-amrwb \
--disable-libopencv \
--disable-libdc1394 \
--disable-libfaac \
--disable-libfreetype \
--disable-libgsm \
--disable-libmp3lame \
--disable-libnut \
--disable-libopenjpeg \
--disable-librtmp \
--disable-libschroedinger \
--disable-libspeex \
--disable-libtheora \
--disable-libvo-aacenc \
--disable-libvo-amrwbenc \
--enable-libvorbis --enable-muxer=ogg --enable-encoder=libvorbis \
--disable-libvpx \
--disable-libx264 \
--disable-libxavs \
--disable-libxvid \
--enable-zlib \
--enable-asm \
--disable-altivec \
$FFMPEG_CPU \
$FFMPEG_FPU \
--disable-vis \
--enable-yasm \
--disable-sram \
--disable-symver
}
post_makeinstall_target() {
rm -rf $INSTALL/usr/share/ffmpeg/examples
}

53
packages/multimedia/ffmpeg/patches/ffmpeg-0001-Support-raw-dvdsub-palette-as-stored-on-normal-dvd-s.patch
View File

@ -1,53 +0,0 @@
From a77624551376c138dcda4d0bce167abe379a0a6b Mon Sep 17 00:00:00 2001
Date: Mon, 18 Feb 2013 11:08:55 +0000
Subject: [PATCH] Support-raw-dvdsub-palette-as-stored-on-normal-dvd-s
---
libavcodec/dvdsubdec.c | 24 ++++++++++++++++++++++++
1 file changed, 24 insertions(+)
diff --git a/libavcodec/dvdsubdec.c b/libavcodec/dvdsubdec.c
index cc9e11b..52ec0a8 100644
--- a/libavcodec/dvdsubdec.c
+++ b/libavcodec/dvdsubdec.c
@@ -54,6 +54,24 @@ static void yuv_a_to_rgba(const uint8_t *ycbcr, const uint8_t *alpha, uint32_t *
}
}
+static void ayvu_to_argb(const uint8_t *ayvu, uint32_t *argb, int num_values)
+{
+ uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
+ uint8_t r, g, b;
+ int i, y, cb, cr, a;
+ int r_add, g_add, b_add;
+
+ for (i = num_values; i > 0; i--) {
+ a = *ayvu++;
+ y = *ayvu++;
+ cr = *ayvu++;
+ cb = *ayvu++;
+ YUV_TO_RGB1_CCIR(cb, cr);
+ YUV_TO_RGB2_CCIR(r, g, b, y);
+ *argb++ = (a << 24) | (r << 16) | (g << 8) | b;
+ }
+}
+
static int decode_run_2bit(GetBitContext *gb, int *color)
{
unsigned int v, t;
@@ -574,6 +592,12 @@ static int dvdsub_init(AVCodecContext *avctx)
if (ctx->palette_str)
parse_palette(ctx, ctx->palette_str);
+
+ if (!ctx->has_palette && avctx->extradata_size == 64) {
+ ayvu_to_argb((uint8_t*)avctx->extradata, ctx->palette, 16);
+ ctx->has_palette = 1;
+ }
+
if (ctx->has_palette) {
int i;
av_log(avctx, AV_LOG_DEBUG, "palette:");
--
1.7.10.4

125
packages/multimedia/ffmpeg/patches/ffmpeg-0002-Change-fallthrough-logic-for-read_seek-to-be-based-o.patch
View File

@ -1,125 +0,0 @@
From bad37718f9fb608ffdd271c96a17f071ad508464 Mon Sep 17 00:00:00 2001
Date: Mon, 18 Feb 2013 11:23:24 +0000
Subject: [PATCH] Change-fallthrough-logic-for-read_seek-to-be-based-o
---
libavformat/asfdec.c | 2 +-
libavformat/gxf.c | 4 ++--
libavformat/jvdec.c | 2 +-
libavformat/oggdec.c | 4 +++-
libavformat/pmpdec.c | 2 +-
libavformat/r3d.c | 2 +-
libavformat/utils.c | 4 +++-
7 files changed, 12 insertions(+), 8 deletions(-)
diff --git a/libavformat/asfdec.c b/libavformat/asfdec.c
index 2dcdf56..2581307 100644
--- a/libavformat/asfdec.c
+++ b/libavformat/asfdec.c
@@ -1415,7 +1415,7 @@ static int asf_read_seek(AVFormatContext *s, int stream_index, int64_t pts, int
AVStream *st = s->streams[stream_index];
if (s->packet_size <= 0)
- return -1;
+ return AVERROR(ENOSYS);
/* Try using the protocol's read_seek if available */
if(s->pb) {
diff --git a/libavformat/gxf.c b/libavformat/gxf.c
index 86e6291..b46e6d6 100644
--- a/libavformat/gxf.c
+++ b/libavformat/gxf.c
@@ -559,7 +559,7 @@ static int gxf_seek(AVFormatContext *s, int stream_index, int64_t timestamp, int
idx = av_index_search_timestamp(st, timestamp - start_time,
AVSEEK_FLAG_ANY | AVSEEK_FLAG_BACKWARD);
if (idx < 0)
- return -1;
+ return AVERROR(ENOSYS);
pos = st->index_entries[idx].pos;
if (idx < st->nb_index_entries - 2)
maxlen = st->index_entries[idx + 2].pos - pos;
@@ -569,7 +569,7 @@ static int gxf_seek(AVFormatContext *s, int stream_index, int64_t timestamp, int
return res;
found = gxf_resync_media(s, maxlen, -1, timestamp);
if (FFABS(found - timestamp) > 4)
- return -1;
+ return AVERROR(ENOSYS);
return 0;
}
diff --git a/libavformat/jvdec.c b/libavformat/jvdec.c
index e941492..8039b2a 100644
--- a/libavformat/jvdec.c
+++ b/libavformat/jvdec.c
@@ -209,7 +209,7 @@ static int read_seek(AVFormatContext *s, int stream_index,
}
if (i < 0 || i >= ast->nb_index_entries)
- return 0;
+ return -1;
if (avio_seek(s->pb, ast->index_entries[i].pos, SEEK_SET) < 0)
return -1;
diff --git a/libavformat/oggdec.c b/libavformat/oggdec.c
index e51db61..7441615 100644
--- a/libavformat/oggdec.c
+++ b/libavformat/oggdec.c
@@ -808,8 +808,10 @@ static int ogg_read_seek(AVFormatContext *s, int stream_index,
ret = ff_seek_frame_binary(s, stream_index, timestamp, flags);
os = ogg->streams + stream_index;
- if (ret < 0)
+ if (ret < 0) {
os->keyframe_seek = 0;
+ ret = AVERROR(ENOSYS);
+ }
return ret;
}
diff --git a/libavformat/pmpdec.c b/libavformat/pmpdec.c
index 2ea37ef..0a89650 100644
--- a/libavformat/pmpdec.c
+++ b/libavformat/pmpdec.c
@@ -161,7 +161,7 @@ static int pmp_seek(AVFormatContext *s, int stream_index, int64_t ts, int flags)
PMPContext *pmp = s->priv_data;
pmp->cur_stream = 0;
// fallback to default seek now
- return -1;
+ return AVERROR(ENOSYS);
}
static int pmp_close(AVFormatContext *s)
diff --git a/libavformat/r3d.c b/libavformat/r3d.c
index 35da81e..452849d 100644
--- a/libavformat/r3d.c
+++ b/libavformat/r3d.c
@@ -366,7 +366,7 @@ static int r3d_seek(AVFormatContext *s, int stream_index, int64_t sample_time, i
int frame_num;
if (!st->avg_frame_rate.num)
- return -1;
+ return AVERROR(ENOSYS);
frame_num = av_rescale_q(sample_time, st->time_base,
av_inv_q(st->avg_frame_rate));
diff --git a/libavformat/utils.c b/libavformat/utils.c
index 5cbf7ca..4b8bcd1 100644
--- a/libavformat/utils.c
+++ b/libavformat/utils.c
@@ -2059,10 +2059,12 @@ static int seek_frame_internal(AVFormatContext *s, int stream_index,
ff_read_frame_flush(s);
ret = s->iformat->read_seek(s, stream_index, timestamp, flags);
} else
- ret = -1;
+ ret = AVERROR(ENOSYS);
if (ret >= 0) {
return 0;
}
+ if (ret != AVERROR(ENOSYS))
+ return ret;
if (s->iformat->read_timestamp && !(s->iformat->flags & AVFMT_NOBINSEARCH)) {
ff_read_frame_flush(s);
--
1.7.10.4

60
packages/multimedia/ffmpeg/patches/ffmpeg-0003-matroska-Check-return-value-of-avio_seek-and-avoid-m.patch
View File

@ -1,60 +0,0 @@
From 6d0c6f74bce43810c2c8a7ca01de5e1f0e7da333 Mon Sep 17 00:00:00 2001
Date: Mon, 18 Feb 2013 11:27:59 +0000
Subject: [PATCH] matroska-Check-return-value-of-avio_seek-and-avoid-m
---
libavformat/matroskadec.c | 19 ++++++++++++-------
1 file changed, 12 insertions(+), 7 deletions(-)
diff --git a/libavformat/matroskadec.c b/libavformat/matroskadec.c
index 87c4a39..d066186 100644
--- a/libavformat/matroskadec.c
+++ b/libavformat/matroskadec.c
@@ -2386,7 +2386,8 @@ static int matroska_read_seek(AVFormatContext *s, int stream_index,
timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
- avio_seek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET);
+ if (avio_seek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET) < 0)
+ return -1;
matroska->current_id = 0;
while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
matroska_clear_queue(matroska);
@@ -2395,16 +2396,11 @@ static int matroska_read_seek(AVFormatContext *s, int stream_index,
}
}
- matroska_clear_queue(matroska);
if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
goto err;
index_min = index;
for (i=0; i < matroska->tracks.nb_elem; i++) {
- tracks[i].audio.pkt_cnt = 0;
- tracks[i].audio.sub_packet_cnt = 0;
- tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
- tracks[i].end_timecode = 0;
if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE
&& tracks[i].stream->discard != AVDISCARD_ALL) {
index_sub = av_index_search_timestamp(tracks[i].stream, st->index_entries[index].timestamp, AVSEEK_FLAG_BACKWARD);
@@ -2415,7 +2411,16 @@ static int matroska_read_seek(AVFormatContext *s, int stream_index,
}
}
- avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
+ if (avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET) < 0)
+ return -1;
+
+ matroska_clear_queue(matroska);
+ for (i=0; i < matroska->tracks.nb_elem; i++) {
+ tracks[i].audio.pkt_cnt = 0;
+ tracks[i].audio.sub_packet_cnt = 0;
+ tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
+ tracks[i].end_timecode = 0;
+ }
matroska->current_id = 0;
st->skip_to_keyframe =
matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
--
1.7.10.4

37
packages/multimedia/ffmpeg/patches/ffmpeg-0004-asf-hacks.patch
View File

@ -1,37 +0,0 @@
From 16ed0150800958b4ca544fd380bda92f21393555 Mon Sep 17 00:00:00 2001
From: Joakim Plate <elupus@ecce.se>
Date: Mon, 12 Sep 2011 21:37:17 +0200
Subject: [PATCH 04/24] asf hacks
---
libavformat/asfdec.c | 11 +++++++++++
1 file changed, 11 insertions(+)
diff --git a/libavformat/asfdec.c b/libavformat/asfdec.c
index a62bab3..7db3bd8 100644
--- a/libavformat/asfdec.c
+++ b/libavformat/asfdec.c
@@ -1277,9 +1277,20 @@ static int asf_read_seek(AVFormatContext *s, int stream_index, int64_t pts, int
ASFContext *asf = s->priv_data;
AVStream *st = s->streams[stream_index];
+ if (pts == 0) {
+ // this is a hack since av_gen_search searches the entire file in this case
+ av_log(s, AV_LOG_DEBUG, "SEEKTO: %"PRId64"\n", s->data_offset);
+ if (avio_seek(s->pb, s->data_offset, SEEK_SET) < 0)
+ return -1;
+ return 0;
+ }
+
if (s->packet_size <= 0)
return AVERROR(ENOSYS);
+ if (st->codec->codec_type != AVMEDIA_TYPE_VIDEO)
+ return -1;
+
/* Try using the protocol's read_seek if available */
if(s->pb) {
int ret = avio_seek_time(s->pb, stream_index, pts, flags);
--
1.7.9.4

24
packages/multimedia/ffmpeg/patches/ffmpeg-0005-if-av_read_packet-returns-AVERROR_IO-we-are-done.-ff.patch
View File

@ -1,24 +0,0 @@
From dac532f278c50002ed2fae5c7e2c5b3edfc4f96e Mon Sep 17 00:00:00 2001
Date: Mon, 18 Feb 2013 11:34:09 +0000
Subject: [PATCH] if-av_read_packet-returns-AVERROR_IO-we-are-done.-ff
---
libavformat/utils.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/libavformat/utils.c b/libavformat/utils.c
index 4b8bcd1..bdb9d17 100644
--- a/libavformat/utils.c
+++ b/libavformat/utils.c
@@ -1386,6 +1386,8 @@ static int read_frame_internal(AVFormatContext *s, AVPacket *pkt)
if (ret < 0) {
if (ret == AVERROR(EAGAIN))
return ret;
+ if (ret == AVERROR(EIO))
+ return ret;
/* flush the parsers */
for(i = 0; i < s->nb_streams; i++) {
st = s->streams[i];
--
1.7.10.4

40
packages/multimedia/ffmpeg/patches/ffmpeg-0006-added-Ticket-7187-TV-Teletext-support-for-DVB-EBU-Te.patch
View File

@ -1,40 +0,0 @@
From 81f4c30d6f4dd4f1da818553c869082b50ab601b Mon Sep 17 00:00:00 2001
Date: Mon, 18 Feb 2013 11:41:56 +0000
Subject: [PATCH] added-Ticket-7187-TV-Teletext-support-for-DVB-EBU-Te
---
libavcodec/avcodec.h | 4 ++++
libavformat/mpegts.c | 2 ++
2 files changed, 6 insertions(+)
diff --git a/libavcodec/avcodec.h b/libavcodec/avcodec.h
index ab7a8a5..fff064c 100644
--- a/libavcodec/avcodec.h
+++ b/libavcodec/avcodec.h
@@ -472,6 +472,10 @@ enum AVCodecID {
AV_CODEC_ID_VPLAYER = MKBETAG('V','P','l','r'),
AV_CODEC_ID_PJS = MKBETAG('P','h','J','S'),
+ /* data codecs */
+ AV_CODEC_ID_VBI_DATA= 0x17500,
+ AV_CODEC_ID_VBI_TELETEXT,
+
/* other specific kind of codecs (generally used for attachments) */
AV_CODEC_ID_FIRST_UNKNOWN = 0x18000, ///< A dummy ID pointing at the start of various fake codecs.
AV_CODEC_ID_TTF = 0x18000,
diff --git a/libavformat/mpegts.c b/libavformat/mpegts.c
index a6e07fa..87269d7 100644
--- a/libavformat/mpegts.c
+++ b/libavformat/mpegts.c
@@ -611,6 +611,8 @@ static const StreamType DESC_types[] = {
{ 0x7b, AVMEDIA_TYPE_AUDIO, AV_CODEC_ID_DTS },
{ 0x56, AVMEDIA_TYPE_SUBTITLE, AV_CODEC_ID_DVB_TELETEXT },
{ 0x59, AVMEDIA_TYPE_SUBTITLE, AV_CODEC_ID_DVB_SUBTITLE }, /* subtitling descriptor */
+ { 0x45, AVMEDIA_TYPE_DATA, AV_CODEC_ID_VBI_DATA }, /* VBI Data descriptor */
+ { 0x46, AVMEDIA_TYPE_DATA, AV_CODEC_ID_VBI_TELETEXT }, /* VBI Teletext descriptor */
{ 0 },
};
--
1.7.10.4

41
packages/multimedia/ffmpeg/patches/ffmpeg-0007-Don-t-accept-mpegts-PMT-that-isn-t-current.patch
View File

@ -1,41 +0,0 @@
From 3bf21b1049abf8991beac79235adfcca9d705eda Mon Sep 17 00:00:00 2001
From: Joakim Plate <elupus@ecce.se>
Date: Sun, 18 Sep 2011 19:16:34 +0200
Subject: [PATCH 07/24] Don't accept mpegts PMT that isn't current
---
libavformat/mpegts.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/libavformat/mpegts.c b/libavformat/mpegts.c
index e7e061e..817fbed 100644
--- a/libavformat/mpegts.c
+++ b/libavformat/mpegts.c
@@ -437,6 +437,7 @@ typedef struct SectionHeader {
uint8_t tid;
uint16_t id;
uint8_t version;
+ uint8_t current;
uint8_t sec_num;
uint8_t last_sec_num;
} SectionHeader;
@@ -508,6 +509,7 @@ static int parse_section_header(SectionHeader *h,
val = get8(pp, p_end);
if (val < 0)
return -1;
+ h->current = val & 0x1;
h->version = (val >> 1) & 0x1f;
val = get8(pp, p_end);
if (val < 0)
@@ -1556,6 +1558,8 @@ static void pat_cb(MpegTSFilter *filter, const uint8_t *section, int section_len
return;
if (h->tid != PAT_TID)
return;
+ if (!h->current)
+ return;
ts->stream->ts_id = h->id;
--
1.7.9.4

43
packages/multimedia/ffmpeg/patches/ffmpeg-0008-Don-t-reparse-PMT-unless-it-s-version-has-changed.patch
View File

@ -1,43 +0,0 @@
From 5fecd135de1486be63e0cef53134c44f8b4becd6 Mon Sep 17 00:00:00 2001
From: Joakim Plate <elupus@ecce.se>
Date: Sun, 18 Sep 2011 19:17:23 +0200
Subject: [PATCH 08/24] Don't reparse PMT unless it's version has changed
---
libavformat/mpegts.c | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/libavformat/mpegts.c b/libavformat/mpegts.c
index 817fbed..8bdafc5 100644
--- a/libavformat/mpegts.c
+++ b/libavformat/mpegts.c
@@ -78,6 +78,7 @@ struct MpegTSFilter {
int pid;
int es_id;
int last_cc; /* last cc code (-1 if first packet) */
+ int last_version; /* last version of data on this pid */
enum MpegTSFilterType type;
union {
MpegTSPESFilter pes_filter;
@@ -334,6 +335,7 @@ static MpegTSFilter *mpegts_open_section_filter(MpegTSContext *ts, unsigned int
filter->pid = pid;
filter->es_id = -1;
filter->last_cc = -1;
+ filter->last_version = -1;
sec = &filter->u.section_filter;
sec->section_cb = section_cb;
sec->opaque = opaque;
@@ -1560,6 +1562,10 @@ static void pat_cb(MpegTSFilter *filter, const uint8_t *section, int section_len
return;
if (!h->current)
return;
+ if (h->version == filter->last_version)
+ return;
+ filter->last_version = h->version;
+ av_dlog(ts->stream, "version=%d\n", filter->last_version);
ts->stream->ts_id = h->id;
--
1.7.9.4

64
packages/multimedia/ffmpeg/patches/ffmpeg-0009-fixed-compile-with-VDPAU-header-versions-without-MPE.patch
View File

@ -1,64 +0,0 @@
From eaa6e4a2d98b3f04b4e1d9cd7db6185015d370d7 Mon Sep 17 00:00:00 2001
Date: Mon, 18 Feb 2013 11:46:38 +0000
Subject: [PATCH] fixed-compile-with-VDPAU-header-versions-without-MPE
---
configure | 2 ++
libavcodec/vdpau.c | 2 ++
libavcodec/vdpau.h | 2 ++
3 files changed, 6 insertions(+)
diff --git a/configure b/configure
index 0707a94..6b9e668 100755
--- a/configure
+++ b/configure
@@ -3748,6 +3748,8 @@ check_header termios.h
check_header unistd.h
check_header vdpau/vdpau.h
check_header vdpau/vdpau_x11.h
+check_cpp_condition vdpau/vdpau.h "defined(VDP_DECODER_PROFILE_MPEG4_PART2_SP)" && enable vdpau_mpeg4_support
+
check_header VideoDecodeAcceleration/VDADecoder.h
check_header windows.h
check_header X11/extensions/XvMClib.h
diff --git a/libavcodec/vdpau.c b/libavcodec/vdpau.c
index 6ac195e..09dcb3e 100644
--- a/libavcodec/vdpau.c
+++ b/libavcodec/vdpau.c
@@ -316,6 +316,7 @@ void ff_vdpau_vc1_decode_picture(MpegEncContext *s, const uint8_t *buf,
render->bitstream_buffers_used = 0;
}
+#ifdef VDP_DECODER_PROFILE_MPEG4_PART2_SP
void ff_vdpau_mpeg4_decode_picture(MpegEncContext *s, const uint8_t *buf,
int buf_size)
{
@@ -369,6 +370,7 @@ void ff_vdpau_mpeg4_decode_picture(MpegEncContext *s, const uint8_t *buf,
ff_draw_horiz_band(s, 0, s->avctx->height);
render->bitstream_buffers_used = 0;
}
+#endif
// Only dummy functions for now
static int vdpau_mpeg2_start_frame(AVCodecContext *avctx, const uint8_t *buffer, uint32_t size)
diff --git a/libavcodec/vdpau.h b/libavcodec/vdpau.h
index 23394b5..67fadb6 100644
--- a/libavcodec/vdpau.h
+++ b/libavcodec/vdpau.h
@@ -56,7 +56,9 @@ union FFVdpPictureInfo {
VdpPictureInfoH264 h264;
VdpPictureInfoMPEG1Or2 mpeg;
VdpPictureInfoVC1 vc1;
- VdpPictureInfoMPEG4Part2 mpeg4;
+#ifdef VDP_DECODER_PROFILE_MPEG4_PART2_SP
+ VdpPictureInfoMPEG4Part2 mpeg4;
+#endif
};
/**
--
1.7.10.4

98
packages/multimedia/ffmpeg/patches/ffmpeg-0010-Read-PID-timestamps-as-well-as-PCR-timestamps-to-fin.patch
View File

@ -1,98 +0,0 @@
From 3c31f5987079c3796f3f7313d97ee3c58fb0d4db Mon Sep 17 00:00:00 2001
Date: Mon, 18 Feb 2013 11:52:01 +0000
Subject: [PATCH] Read-PID-timestamps-as-well-as-PCR-timestamps-to-fin
---
libavformat/mpegts.c | 48 ++++++++++++++++++++++++++++++++++++++++++++++--
1 file changed, 46 insertions(+), 2 deletions(-)
diff --git a/libavformat/mpegts.c b/libavformat/mpegts.c
index 3befaf8..21df7b3 100644
--- a/libavformat/mpegts.c
+++ b/libavformat/mpegts.c
@@ -1985,6 +1985,44 @@ static int parse_pcr(int64_t *ppcr_high, int *ppcr_low,
return 0;
}
+static int parse_timestamp(int64_t *ts, const uint8_t *buf)
+{
+ int afc, flags;
+ const uint8_t *p;
+
+ if(!(buf[1] & 0x40)) /* must be a start packet */
+ return -1;
+
+ afc = (buf[3] >> 4) & 3;
+ p = buf + 4;
+ if (afc == 0 || afc == 2) /* invalid or only adaption field */
+ return -1;
+ if (afc == 3)
+ p += p[0] + 1;
+ if (p >= buf + TS_PACKET_SIZE)
+ return -1;
+
+ if (p[0] != 0x00 || p[1] != 0x00 || p[2] != 0x01) /* packet_start_code_prefix */
+ return -1;
+
+ flags = p[3] | 0x100; /* stream type */
+ if (!((flags >= 0x1c0 && flags <= 0x1df) ||
+ (flags >= 0x1e0 && flags <= 0x1ef) ||
+ (flags == 0x1bd) || (flags == 0x1fd)))
+ return -1;
+
+ flags = p[7];
+ if ((flags & 0xc0) == 0x80) {
+ *ts = ff_parse_pes_pts(p+9);
+ return 0;
+ } else if ((flags & 0xc0) == 0xc0) {
+ *ts = ff_parse_pes_pts(p+9+5);
+ return 0;
+ }
+ return -1;
+}
+
+
static int mpegts_read_header(AVFormatContext *s)
{
MpegTSContext *ts = s->priv_data;
@@ -2176,6 +2214,7 @@ static av_unused int64_t mpegts_get_pcr(AVFormatContext *s, int stream_index,
int64_t pos, timestamp;
uint8_t buf[TS_PACKET_SIZE];
int pcr_l, pcr_pid = ((PESContext*)s->streams[stream_index]->priv_data)->pcr_pid;
+ int pid = ((PESContext*)s->streams[stream_index]->priv_data)->pid;
pos = ((*ppos + ts->raw_packet_size - 1 - ts->pos47) / ts->raw_packet_size) * ts->raw_packet_size + ts->pos47;
while(pos < pos_limit) {
if (avio_seek(s->pb, pos, SEEK_SET) < 0)
@@ -2193,6 +2232,11 @@ static av_unused int64_t mpegts_get_pcr(AVFormatContext *s, int stream_index,
*ppos = pos;
return timestamp;
}
+ if ((pid < 0 || (AV_RB16(buf + 1) & 0x1fff) == pid) &&
+ parse_timestamp(&timestamp, buf) == 0) {
+ *ppos = pos;
+ return timestamp;
+ }
pos += ts->raw_packet_size;
}
@@ -2294,7 +2338,7 @@ AVInputFormat ff_mpegts_demuxer = {
.read_header = mpegts_read_header,
.read_packet = mpegts_read_packet,
.read_close = mpegts_read_close,
- .read_timestamp = mpegts_get_dts,
+ .read_timestamp = mpegts_get_pcr,
.flags = AVFMT_SHOW_IDS | AVFMT_TS_DISCONT,
};
@@ -2305,7 +2349,7 @@ AVInputFormat ff_mpegtsraw_demuxer = {
.read_header = mpegts_read_header,
.read_packet = mpegts_raw_read_packet,
.read_close = mpegts_read_close,
- .read_timestamp = mpegts_get_dts,
+ .read_timestamp = mpegts_get_pcr,
.flags = AVFMT_SHOW_IDS | AVFMT_TS_DISCONT,
.priv_class = &mpegtsraw_class,
};
--
1.7.10.4

68
packages/multimedia/ffmpeg/patches/ffmpeg-0011-Get-stream-durations-using-read_timestamp.patch
View File

@ -1,68 +0,0 @@
From f3bdfee28332b1ccc794186d91a1f9e082004a1e Mon Sep 17 00:00:00 2001
Date: Mon, 18 Feb 2013 11:55:37 +0000
Subject: [PATCH] Get-stream-durations-using-read_timestamp
---
libavformat/utils.c | 39 +++++++++++++++++++++++++++++++++++++++
1 file changed, 39 insertions(+)
diff --git a/libavformat/utils.c b/libavformat/utils.c
index bdb9d17..3e7dec9 100644
--- a/libavformat/utils.c
+++ b/libavformat/utils.c
@@ -2283,6 +2283,41 @@ static void estimate_timings_from_bit_rate(AVFormatContext *ic)
#define DURATION_MAX_READ_SIZE 250000LL
#define DURATION_MAX_RETRY 4
+static void av_estimate_timings_from_pts2(AVFormatContext *ic, int64_t old_offset)
+{
+ AVStream *st;
+ int i, step= 1024;
+ int64_t ts, pos;
+
+ for(i=0;i<ic->nb_streams;i++) {
+ st = ic->streams[i];
+
+ pos = 0;
+ ts = ic->iformat->read_timestamp(ic, i, &pos, DURATION_MAX_READ_SIZE);
+ if (ts == AV_NOPTS_VALUE)
+ continue;
+ if (st->start_time > ts || st->start_time == AV_NOPTS_VALUE)
+ st->start_time = ts;
+
+ pos = avio_size(ic->pb) - 1;
+ do {
+ pos -= step;
+ ts = ic->iformat->read_timestamp(ic, i, &pos, pos + step);
+ step += step;
+ } while (ts == AV_NOPTS_VALUE && pos >= step && step < DURATION_MAX_READ_SIZE);
+
+ if (ts == AV_NOPTS_VALUE)
+ continue;
+
+ if (st->duration < ts - st->start_time || st->duration == AV_NOPTS_VALUE)
+ st->duration = ts - st->start_time;
+ }
+
+ fill_all_stream_timings(ic);
+
+ avio_seek(ic->pb, old_offset, SEEK_SET);
+}
+
/* only usable for MPEG-PS streams */
static void estimate_timings_from_pts(AVFormatContext *ic, int64_t old_offset)
{
@@ -2382,6 +2417,10 @@ static void estimate_timings(AVFormatContext *ic, int64_t old_offset)
the components */
fill_all_stream_timings(ic);
ic->duration_estimation_method = AVFMT_DURATION_FROM_STREAM;
+ } else if (ic->iformat->read_timestamp &&
+ file_size && ic->pb->seekable) {
+ /* get accurate estimate from the PTSes */
+ av_estimate_timings_from_pts2(ic, old_offset);
} else {
av_log(ic, AV_LOG_WARNING, "Estimating duration from bitrate, this may be inaccurate\n");
/* less precise: use bitrate info */
--
1.7.10.4

33
packages/multimedia/ffmpeg/patches/ffmpeg-0013-aacenc-add-recognized-profiles-array.patch
View File

@ -1,33 +0,0 @@
From 51e09f5022d1250802cbec8017a2c5e76160bdcb Mon Sep 17 00:00:00 2001
Date: Mon, 18 Feb 2013 16:53:54 +0000
Subject: [PATCH] aacenc-add-recognized-profiles-array
---
libavcodec/aacenc.c | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/libavcodec/aacenc.c b/libavcodec/aacenc.c
index 348d3a2..3cbca03 100644
--- a/libavcodec/aacenc.c
+++ b/libavcodec/aacenc.c
@@ -818,6 +818,11 @@ static const int mpeg4audio_sample_rates[16] = {
24000, 22050, 16000, 12000, 11025, 8000, 7350
};
+static const AVProfile profiles[] = {
+ { FF_PROFILE_AAC_LOW, "Low" },
+ { FF_PROFILE_UNKNOWN },
+};
+
AVCodec ff_aac_encoder = {
.name = "aac",
.type = AVMEDIA_TYPE_AUDIO,
@@ -833,4 +838,5 @@ AVCodec ff_aac_encoder = {
AV_SAMPLE_FMT_NONE },
.long_name = NULL_IF_CONFIG_SMALL("AAC (Advanced Audio Coding)"),
.priv_class = &aacenc_class,
+ .profiles = profiles,
};
--
1.7.10.4

28
packages/multimedia/ffmpeg/patches/ffmpeg-0014-changed-allow-4-second-skew-between-streams-in-mov-b.patch
View File

@ -1,28 +0,0 @@
From 5cc735503bfe1242606f19638bffeb686dd4d1b5 Mon Sep 17 00:00:00 2001
From: Joakim Plate <elupus@ecce.se>
Date: Wed, 8 Dec 2010 14:03:43 +0000
Subject: [PATCH 14/24] changed: allow 4 second skew between streams in mov
before attempting to seek
---
libavformat/mov.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/libavformat/mov.c b/libavformat/mov.c
index 9da8eab..0f9836c 100644
--- a/libavformat/mov.c
+++ b/libavformat/mov.c
@@ -2763,8 +2763,8 @@ static AVIndexEntry *mov_find_next_sample(AVFormatContext *s, AVStream **st)
if (!sample || (!s->pb->seekable && current_sample->pos < sample->pos) ||
(s->pb->seekable &&
((msc->pb != s->pb && dts < best_dts) || (msc->pb == s->pb &&
- ((FFABS(best_dts - dts) <= AV_TIME_BASE && current_sample->pos < sample->pos) ||
- (FFABS(best_dts - dts) > AV_TIME_BASE && dts < best_dts)))))) {
+ ((FFABS(best_dts - dts) <= 4*AV_TIME_BASE && current_sample->pos < sample->pos) ||
+ (FFABS(best_dts - dts) > 4*AV_TIME_BASE && dts < best_dts)))))) {
sample = current_sample;
best_dts = dts;
*st = avst;
--
1.7.9.4

38
packages/multimedia/ffmpeg/patches/ffmpeg-0015-fixed-memleak-in-mpegts-demuxer-on-some-malformed-mp.patch
View File

@ -1,38 +0,0 @@
From b83c9a2505338cdf021dd499c26686e82bcbc066 Mon Sep 17 00:00:00 2001
From: Joakim Plate <elupus@ecce.se>
Date: Fri, 26 Nov 2010 20:56:48 +0000
Subject: [PATCH 15/24] fixed: memleak in mpegts demuxer on some malformed
(??) mpegts files with too large pes packets
at-visions sample file brokenStream.mpg
---
libavformat/mpegts.c | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/libavformat/mpegts.c b/libavformat/mpegts.c
index ba2f163..c374cb9 100644
--- a/libavformat/mpegts.c
+++ b/libavformat/mpegts.c
@@ -664,6 +664,10 @@ static int mpegts_set_stream_info(AVStream *st, PESContext *pes,
static void new_pes_packet(PESContext *pes, AVPacket *pkt)
{
+ if(pkt->data) {
+ av_log(pes->stream, AV_LOG_ERROR, "ignoring previously allocated packet on stream %d\n", pkt->stream_index);
+ av_free_packet(pkt);
+ }
av_init_packet(pkt);
pkt->destruct = av_destruct_packet;
@@ -2117,6 +2121,8 @@ static int mpegts_read_packet(AVFormatContext *s,
int ret, i;
ts->pkt = pkt;
+ ts->pkt->data = NULL;
+
ret = handle_packets(ts, 0);
if (ret < 0) {
/* flush pes data left */
--
1.7.9.4

37
packages/multimedia/ffmpeg/patches/ffmpeg-0016-Speed-up-mpegts-av_find_stream_info.patch
View File

@ -1,37 +0,0 @@
From c240357f903f20ba51001fb90aa2de415d5be825 Mon Sep 17 00:00:00 2001
From: Joakim Plate <elupus@ecce.se>
Date: Mon, 28 Jun 2010 21:26:54 +0000
Subject: [PATCH 16/24] Speed up mpegts av_find_stream_info
---
libavformat/mpegts.c | 7 +++++--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/libavformat/mpegts.c b/libavformat/mpegts.c
index c374cb9..6da6db5 100644
--- a/libavformat/mpegts.c
+++ b/libavformat/mpegts.c
@@ -825,7 +825,7 @@ static int mpegts_push_data(MpegTSFilter *filter,
goto skip;
/* stream not present in PMT */
- if (!pes->st) {
+ if (ts->auto_guess && !pes->st) {
pes->st = avformat_new_stream(ts->stream, NULL);
if (!pes->st)
return AVERROR(ENOMEM);
@@ -2013,7 +2013,10 @@ static int mpegts_read_header(AVFormatContext *s,
av_dlog(ts->stream, "tuning done\n");
- s->ctx_flags |= AVFMTCTX_NOHEADER;
+ /* only flag NOHEADER if we are in file mode,
+ in streaming mode scanning may take too long for users */
+ if (pb->seekable)
+ s->ctx_flags |= AVFMTCTX_NOHEADER;
} else {
AVStream *st;
int pcr_pid, pid, nb_packets, nb_pcrs, ret, pcr_l;
--
1.7.9.4

40
packages/multimedia/ffmpeg/patches/ffmpeg-0017-allow-customizing-shared-library-soname-name-with-ma.patch
View File

@ -1,40 +0,0 @@
From 4f4aacd3d85c5caecbc0b0f02739aff0b9d8f657 Mon Sep 17 00:00:00 2001
Date: Mon, 18 Feb 2013 16:58:33 +0000
Subject: [PATCH] allow-customizing-shared-library-soname-name-with-ma
---
configure | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/configure b/configure
index 6b9e668..d4aa8f2 100755
--- a/configure
+++ b/configure
@@ -269,6 +269,7 @@ Advanced options (experts only):
--arch=ARCH select architecture [$arch]
--cpu=CPU select the minimum required CPU (affects
instruction selection, may crash on older CPUs)
+ --custom-libname-with-major=NAME custom library name with major [$SLIBNAME_WITH_MAJOR]
--enable-pic build position-independent code
--enable-sram allow use of on-chip SRAM
--enable-thumb compile for Thumb instruction set
@@ -1527,6 +1528,7 @@ CMDLINE_SET="
cpu
cross_prefix
cxx
+ custom_libname_with_major
dep_cc
extra_version
host_cc
@@ -3408,6 +3410,8 @@ EOF
exit 1;
fi
+test -n "$custom_libname_with_major" && SLIBNAME_WITH_MAJOR="$custom_libname_with_major"
+
die_license_disabled() {
enabled $1 || { enabled $2 && die "$2 is $1 and --enable-$1 is not specified."; }
}
--
1.7.10.4

75
packages/multimedia/ffmpeg/patches/ffmpeg-0018-dxva-mpeg2-Allocate-slices-array-dynamically-fixes-v.patch
View File

@ -1,75 +0,0 @@
From 40f4c15370f7027dc5422edcb10e8a3b7e58e83d Mon Sep 17 00:00:00 2001
From: CrystalP <CrystalP@xbmc.org>
Date: Wed, 5 Oct 2011 12:38:30 -0400
Subject: [PATCH 18/24] dxva-mpeg2 Allocate slices array dynamically - fixes
videos with > 175 slices. They used to result in
images with a black bottom.
sample on team ftp samples/PR471/too_many_slices.ts
Inspired by the vaapi code to reallocate the slices array for each new slice.
Could be more efficient if the array could be preserved for all frames and
freed only at the end of the video, but there doesn't seem to be anywhere
appropriate to free the memory at the end.
Alternative is to allocate the proper size straight away for a new frame,
instead of realloc'ing for each slice.
---
libavcodec/dxva2_mpeg2.c | 14 +++++++++++---
1 file changed, 11 insertions(+), 3 deletions(-)
diff --git a/libavcodec/dxva2_mpeg2.c b/libavcodec/dxva2_mpeg2.c
index 951305d..8ba83b6 100644
--- a/libavcodec/dxva2_mpeg2.c
+++ b/libavcodec/dxva2_mpeg2.c
@@ -22,12 +22,12 @@
#include "dxva2_internal.h"
-#define MAX_SLICES (SLICE_MAX_START_CODE - SLICE_MIN_START_CODE + 1)
struct dxva2_picture_context {
DXVA_PictureParameters pp;
DXVA_QmatrixData qm;
unsigned slice_count;
- DXVA_SliceInfo slice[MAX_SLICES];
+ DXVA_SliceInfo *slice;
+ unsigned int slice_alloc;
const uint8_t *bitstream;
unsigned bitstream_size;
@@ -220,6 +220,8 @@ static int start_frame(AVCodecContext *avctx,
fill_quantization_matrices(avctx, ctx, s, &ctx_pic->qm);
ctx_pic->slice_count = 0;
+ ctx_pic->slice = NULL;
+ ctx_pic->slice_alloc = 0;
ctx_pic->bitstream_size = 0;
ctx_pic->bitstream = NULL;
return 0;
@@ -232,9 +234,14 @@ static int decode_slice(AVCodecContext *avctx,
struct dxva2_picture_context *ctx_pic =
s->current_picture_ptr->f.hwaccel_picture_private;
unsigned position;
+ DXVA_SliceInfo* slice;
- if (ctx_pic->slice_count >= MAX_SLICES)
+ slice = av_fast_realloc(ctx_pic->slice,
+ &ctx_pic->slice_alloc,
+ (ctx_pic->slice_count + 1) * sizeof(DXVA_SliceInfo));
+ if (!slice)
return -1;
+ ctx_pic->slice = slice;
if (!ctx_pic->bitstream)
ctx_pic->bitstream = buffer;
@@ -258,6 +265,7 @@ static int end_frame(AVCodecContext *avctx)
if (!ret)
ff_mpeg_draw_horiz_band(s, 0, avctx->height);
return ret;
+ av_freep(ctx_pic->slice);
}
AVHWAccel ff_mpeg2_dxva2_hwaccel = {
--
1.7.9.4

75
packages/multimedia/ffmpeg/patches/ffmpeg-0019-dxva-mpeg2-speed-up-slice-allocation.patch
View File

@ -1,75 +0,0 @@
From 681f74b224e16a4df7f8c4e31a9be56975d57e10 Mon Sep 17 00:00:00 2001
From: CrystalP <CrystalP@xbmc.org>
Date: Mon, 10 Oct 2011 19:42:50 -0400
Subject: [PATCH 19/24] dxva-mpeg2 speed up slice allocation
The number of slices is not very likely to change from frame to frame, so
at the beginning of a new frame, allocate memory for the amount of slices of
the previous frame. Saves a lot of reallocation, for some TV capture samples
there are over 200 slices.
There wasn't anywhere really appropriate to store last_slice_count (needs to
live from first frame to last frame), so this is likely to cause discussion to
merge upstream.
Adding members to dxva_context breaks ABI, which we don't care too much about
since on Windows we don't support external ffmpeg.
dxva mpeg2 code also has access to MpegEncContext, but adding there would
likely break ABI as well.
---
libavcodec/dxva2.h | 1 +
libavcodec/dxva2_mpeg2.c | 12 ++++++++++++
2 files changed, 13 insertions(+)
diff --git a/libavcodec/dxva2.h b/libavcodec/dxva2.h
index fc99560..16a6994 100644
--- a/libavcodec/dxva2.h
+++ b/libavcodec/dxva2.h
@@ -66,6 +66,7 @@ struct dxva_context {
* Private to the FFmpeg AVHWAccel implementation
*/
unsigned report_id;
+ unsigned last_slice_count;
};
#endif /* AVCODEC_DXVA_H */
diff --git a/libavcodec/dxva2_mpeg2.c b/libavcodec/dxva2_mpeg2.c
index 8ba83b6..90507f9 100644
--- a/libavcodec/dxva2_mpeg2.c
+++ b/libavcodec/dxva2_mpeg2.c
@@ -222,6 +222,16 @@
ctx_pic->slice_count = 0;
ctx_pic->bitstream_size = 0;
ctx_pic->bitstream = NULL;
+
+ if (ctx->last_slice_count > 0)
+ {
+ ctx_pic->slice = av_fast_realloc(NULL,
+ &ctx_pic->slice_alloc,
+ ctx->last_slice_count * sizeof(DXVA_SliceInfo));
+ if (!ctx_pic->slice)
+ return -1;
+ }
+
return 0;
}
@@ -251,6 +261,7 @@
struct MpegEncContext *s = avctx->priv_data;
struct dxva2_picture_context *ctx_pic =
s->current_picture_ptr->f.hwaccel_picture_private;
+ struct dxva_context *ctx = avctx->hwaccel_context;
int ret;
if (ctx_pic->slice_count <= 0 || ctx_pic->bitstream_size <= 0)
@@ -261,6 +272,8 @@
commit_bitstream_and_slice_buffer);
if (!ret)
ff_mpeg_draw_horiz_band(s, 0, avctx->height);
+
+ ctx->last_slice_count = ctx_pic->slice_count;
return ret;
}
--
1.7.9.4

46
packages/multimedia/ffmpeg/patches/ffmpeg-0020-dxva-vc1-Take-BI-into-account-for-forward-and-backwa.patch
View File

@ -1,46 +0,0 @@
From 85a109a59218337bbfac6b3ff54e37499dcbb306 Mon Sep 17 00:00:00 2001
Date: Mon, 18 Feb 2013 17:06:15 +0000
Subject: [PATCH] dxva-vc1-Take-BI-into-account-for-forward-and-backwa
---
libavcodec/dxva2_vc1.c | 10 ++++++----
1 file changed, 6 insertions(+), 4 deletions(-)
diff --git a/libavcodec/dxva2_vc1.c b/libavcodec/dxva2_vc1.c
index ed4836d..26e1ad4 100644
--- a/libavcodec/dxva2_vc1.c
+++ b/libavcodec/dxva2_vc1.c
@@ -38,15 +38,17 @@ static void fill_picture_parameters(AVCodecContext *avctx,
{
const MpegEncContext *s = &v->s;
const Picture *current_picture = s->current_picture_ptr;
+ BYTE bPicIntra = s->pict_type == AV_PICTURE_TYPE_I || v->bi_type == 1;
+ BYTE bPicBackwardPrediction = s->pict_type == AV_PICTURE_TYPE_B && v->bi_type == 0;
memset(pp, 0, sizeof(*pp));
pp->wDecodedPictureIndex =
pp->wDeblockedPictureIndex = ff_dxva2_get_surface_index(ctx, current_picture);
- if (s->pict_type != AV_PICTURE_TYPE_I && !v->bi_type)
+ if (!bPicIntra)
pp->wForwardRefPictureIndex = ff_dxva2_get_surface_index(ctx, &s->last_picture);
else
pp->wForwardRefPictureIndex = 0xffff;
- if (s->pict_type == AV_PICTURE_TYPE_B && !v->bi_type)
+ if (bPicBackwardPrediction)
pp->wBackwardRefPictureIndex = ff_dxva2_get_surface_index(ctx, &s->next_picture);
else
pp->wBackwardRefPictureIndex = 0xffff;
@@ -69,8 +71,8 @@ static void fill_picture_parameters(AVCodecContext *avctx,
if (s->picture_structure & PICT_BOTTOM_FIELD)
pp->bPicStructure |= 0x02;
pp->bSecondField = v->interlace && v->fcm == ILACE_FIELD && v->second_field;
- pp->bPicIntra = s->pict_type == AV_PICTURE_TYPE_I || v->bi_type;
- pp->bPicBackwardPrediction = s->pict_type == AV_PICTURE_TYPE_B && !v->bi_type;
+ pp->bPicIntra = bPicIntra;
+ pp->bPicBackwardPrediction = bPicBackwardPrediction;
pp->bBidirectionalAveragingMode = (1 << 7) |
((ctx->cfg->ConfigIntraResidUnsigned != 0) << 6) |
((ctx->cfg->ConfigResidDiffAccelerator != 0) << 5) |
--
1.7.10.4

26
packages/multimedia/ffmpeg/patches/ffmpeg-0021-dxva-vc1-Pass-overlapping-transforms-hint.patch
View File

@ -1,26 +0,0 @@
From 484c01de4d5dba595ad8dc5582d4f8d2b7b7f2d4 Mon Sep 17 00:00:00 2001
Date: Mon, 18 Feb 2013 17:11:04 +0000
Subject: [PATCH] dxva-vc1-Pass-overlapping-transforms-hint
---
libavcodec/dxva2_vc1.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/libavcodec/dxva2_vc1.c b/libavcodec/dxva2_vc1.c
index 26e1ad4..3dc32cb 100644
--- a/libavcodec/dxva2_vc1.c
+++ b/libavcodec/dxva2_vc1.c
@@ -103,7 +103,9 @@ static void fill_picture_parameters(AVCodecContext *avctx,
(v->rangered << 3) |
(s->max_b_frames );
pp->bPicExtrapolation = (!v->interlace || v->fcm == PROGRESSIVE) ? 1 : 2;
- pp->bPicDeblocked = ((!pp->bPicBackwardPrediction && v->overlap) << 6) |
+ pp->bPicDeblocked = ((v->overlap == 1 &&
+ pp->bPicBackwardPrediction == 0 &&
+ ctx->cfg->ConfigResidDiffHost == 0) << 6) |
((v->profile != PROFILE_ADVANCED && v->rangeredfrm) << 5) |
(s->loop_filter << 1);
pp->bPicDeblockConfined = (v->postprocflag << 7) |
--
1.7.10.4

67
packages/multimedia/ffmpeg/patches/ffmpeg-0022-dxva-h264-Fix-dxva-playback-of-streams-that-don-t-st.patch
View File

@ -1,67 +0,0 @@
From 95b4c0af771121001967e01c1a13125c13178e25 Mon Sep 17 00:00:00 2001
From: marc <mhocking@ubuntu-desktop.(none)>
Date: Mon, 18 Feb 2013 17:18:18 +0000
Subject: [PATCH] dxva-h264 Fix dxva playback of streams that don't start
with an I-Frame.
---
libavcodec/dxva2_h264.c | 8 ++++++++
libavcodec/h264.c | 2 ++
libavcodec/h264.h | 2 ++
3 files changed, 12 insertions(+)
diff --git a/libavcodec/dxva2_h264.c b/libavcodec/dxva2_h264.c
index f7d4e5d..2ad4afd 100644
--- a/libavcodec/dxva2_h264.c
+++ b/libavcodec/dxva2_h264.c
@@ -431,6 +431,14 @@
if (ctx_pic->slice_count <= 0 || ctx_pic->bitstream_size <= 0)
return -1;
+
+ // Wait for an I-frame before start decoding. Workaround for ATI UVD and UVD+ GPUs
+ if (!h->got_first_iframe) {
+ if (!(ctx_pic->pp.wBitFields & (1 << 15)))
+ return -1;
+ h->got_first_iframe = 1;
+ }
+
ret = ff_dxva2_common_end_frame(avctx, h->cur_pic_ptr,
&ctx_pic->pp, sizeof(ctx_pic->pp),
&ctx_pic->qm, sizeof(ctx_pic->qm),
diff --git a/libavcodec/h264.c b/libavcodec/h264.c
index 0aab4e7..aee9407 100644
--- a/libavcodec/h264.c
+++ b/libavcodec/h264.c
@@ -2183,6 +2183,7 @@ static void flush_dpb(AVCodecContext *avctx)
h->delayed_pic[i]->f.reference = 0;
h->delayed_pic[i] = NULL;
}
+ h->got_first_iframe = 0;
flush_change(h);
ff_mpeg_flush(avctx);
@@ -2597,6 +2598,7 @@ static int h264_slice_header_init(H264Context *h, int reinit)
}
s->first_field = 0;
h->prev_interlaced_frame = 1;
+ h->got_first_iframe = 0;
init_scan_tables(h);
if (ff_h264_alloc_tables(h) < 0) {
diff --git a/libavcodec/h264.h b/libavcodec/h264.h
index 4b6a19e..06c2384 100644
--- a/libavcodec/h264.h
+++ b/libavcodec/h264.h
@@ -588,6 +588,8 @@ typedef struct H264Context {
int luma_weight_flag[2]; ///< 7.4.3.2 luma_weight_lX_flag
int chroma_weight_flag[2]; ///< 7.4.3.2 chroma_weight_lX_flag
+ int got_first_iframe;
+
// Timestamp stuff
int sei_buffering_period_present; ///< Buffering period SEI flag
int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
--
1.7.10.4

44
packages/multimedia/ffmpeg/patches/ffmpeg-0024-add-public-version-of-ff_read_frame_flush.patch
View File

@ -1,44 +0,0 @@
From f9cbb94f63aaaf889c21445e71fdd258f9ebcad1 Mon Sep 17 00:00:00 2001
Date: Mon, 18 Feb 2013 17:31:52 +0000
Subject: [PATCH] add-public-version-of-ff_read_frame_flush
---
libavformat/avformat.h | 5 +++++
libavformat/utils.c | 5 +++++
2 files changed, 10 insertions(+)
diff --git a/libavformat/avformat.h b/libavformat/avformat.h
index c907d4e..c0f517c 100644
--- a/libavformat/avformat.h
+++ b/libavformat/avformat.h
@@ -1615,6 +1615,11 @@ int av_read_packet(AVFormatContext *s, AVPacket *pkt);
int av_read_frame(AVFormatContext *s, AVPacket *pkt);
/**
+ * Clear out any buffered data in context
+ */
+void av_read_frame_flush(AVFormatContext *s);
+
+/**
* Seek to the keyframe at timestamp.
* 'timestamp' in 'stream_index'.
* @param stream_index If stream_index is (-1), a default
diff --git a/libavformat/utils.c b/libavformat/utils.c
index 3e7dec9..72925bb 100644
--- a/libavformat/utils.c
+++ b/libavformat/utils.c
@@ -1636,6 +1636,11 @@ void ff_read_frame_flush(AVFormatContext *s)
}
}
+void av_read_frame_flush(AVFormatContext *s)
+{
+ ff_read_frame_flush(s);
+}
+
void ff_update_cur_dts(AVFormatContext *s, AVStream *ref_st, int64_t timestamp)
{
int i;
--
1.7.10.4

51
packages/multimedia/ffmpeg/patches/ffmpeg-0026-Handle-return-value-of-BeginFrame-better.patch
View File

@ -1,51 +0,0 @@
From d659958d58dfec08f4666a8add325c8154a662de Mon Sep 17 00:00:00 2001
From: CrystalP <CrystalP@xbmc.org>
Date: Fri, 11 Nov 2011 19:10:54 -0500
Subject: [PATCH 1/1] Handle return value of BeginFrame better.
The nVidia cards sometimes return E_PENDING and need time before they can start
decoding a new frame.
Helps nVidia cards with blocky pictures/pixellation artifacts after skip or
when CPU is busy.
---
libavcodec/dxva2.c | 14 ++++++++++++--
1 file changed, 12 insertions(+), 2 deletions(-)
diff --git a/libavcodec/dxva2.c b/libavcodec/dxva2.c
index b6f8aea..3a10f69 100644
--- a/libavcodec/dxva2.c
+++ b/libavcodec/dxva2.c
@@ -21,6 +21,9 @@
*/
#include "dxva2_internal.h"
+#include <unistd.h>
+
+#define MAX_RETRY_ON_PENDING 50
void *ff_dxva2_get_surface(const Picture *picture)
{
@@ -88,10 +91,17 @@ int ff_dxva2_common_end_frame(AVCodecContext *avctx, MpegEncContext *s,
DXVA2_DecodeBufferDesc buffer[4];
DXVA2_DecodeExecuteParams exec = { 0 };
int result;
+ HRESULT hr;
+ int tries = 0;
- if (FAILED(IDirectXVideoDecoder_BeginFrame(ctx->decoder,
+ while ((hr=IDirectXVideoDecoder_BeginFrame(ctx->decoder,
ff_dxva2_get_surface(pic),
- NULL))) {
+ NULL)) == E_PENDING
+ && tries < MAX_RETRY_ON_PENDING) {
+ usleep(1000);
+ tries++;
+ }
+ if (FAILED(hr)) {
av_log(avctx, AV_LOG_ERROR, "Failed to begin frame\n");
return -1;
}
--
1.7.9.4

26
packages/multimedia/ffmpeg/patches/ffmpeg-0028-ffmpeg-1.2-fixed-dvd-still-frames-ended-up-in-intern.patch
View File

@ -1,26 +0,0 @@
From a356770e3a73d79c9e4a451366930f557e3a2fa6 Mon Sep 17 00:00:00 2001
From: Voyager1 <voyager@xbmc.org>
Date: Sat, 13 Apr 2013 15:38:10 +0200
Subject: [PATCH 4/4] ffmpeg 1.2 - fixed dvd still frames ended up in internal
lavf
---
libavformat/utils.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/libavformat/utils.c b/libavformat/utils.c
index c3bd3d9..72f59d2 100644
--- a/libavformat/utils.c
+++ b/libavformat/utils.c
@@ -706,7 +706,7 @@ no_packet:
if(end || av_log2(pd->buf_size) != av_log2(pd->buf_size - pkt->size)){
int score= set_codec_from_probe_data(s, st, pd);
- if( (st->codec->codec_id != AV_CODEC_ID_NONE && score > AVPROBE_SCORE_RETRY)
+ if( (st->codec->codec_id != AV_CODEC_ID_NONE && score > AVPROBE_SCORE_RETRY-1)
|| end){
pd->buf_size=0;
av_freep(&pd->buf);
--
1.8.0.msysgit.0

24
packages/multimedia/ffmpeg/patches/ffmpeg-0029-stop-forcing-gas-preprocessor-for-darwin.patch
View File

@ -1,24 +0,0 @@
From bc85fb8d204c7090dc6856d880f7556e7291d353 Mon Sep 17 00:00:00 2001
Date: Mon, 18 Feb 2013 21:18:39 +0000
Subject: [PATCH] stop-forcing-gas-preprocessor-for-darwin
---
configure | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/configure b/configure
index d4aa8f2..980346e 100755
--- a/configure
+++ b/configure
@@ -3166,7 +3166,7 @@ case $target_os in
strip="strip -d"
;;
darwin)
- gas="gas-preprocessor.pl $cc"
+ #gas="gas-preprocessor.pl $cc"
enabled ppc && add_asflags -force_cpusubtype_ALL
SHFLAGS='-dynamiclib -Wl,-single_module -Wl,-install_name,$(SHLIBDIR)/$(SLIBNAME_WITH_MAJOR),-current_version,$(LIBVERSION),-compatibility_version,$(LIBMAJOR)'
enabled x86_32 && append SHFLAGS -Wl,-read_only_relocs,suppress
--
1.7.10.4

32
packages/multimedia/ffmpeg/patches/ffmpeg-0036-backport-register-vdpau-hwaccel-for-mpeg12-fe1f36547d0be963e352de0cde1a6cba59ea2e78.patch
View File

@ -1,32 +0,0 @@
From bb6ba57092c402b6f2e5edf6d1691beafafa0460 Mon Sep 17 00:00:00 2001
From: Rainer Hochecker <fernetmenta@online.de>
Date: Mon, 6 May 2013 20:58:28 +0200
Subject: [PATCH] ffmpeg backport: register vdpau hwaccel for mpeg12
---
libavcodec/mpeg12.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/libavcodec/mpeg12.c b/libavcodec/mpeg12.c
index 9d2743a..0f94772 100644
--- a/libavcodec/mpeg12.c
+++ b/libavcodec/mpeg12.c
@@ -1202,6 +1202,7 @@ static const enum AVPixelFormat mpeg1_hwaccel_pixfmt_list_420[] = {
#endif
#if CONFIG_MPEG1_VDPAU_HWACCEL
AV_PIX_FMT_VDPAU_MPEG1,
+ AV_PIX_FMT_VDPAU,
#endif
AV_PIX_FMT_YUV420P,
AV_PIX_FMT_NONE
@@ -1214,6 +1215,7 @@ static const enum AVPixelFormat mpeg2_hwaccel_pixfmt_list_420[] = {
#endif
#if CONFIG_MPEG2_VDPAU_HWACCEL
AV_PIX_FMT_VDPAU_MPEG2,
+ AV_PIX_FMT_VDPAU,
#endif
#if CONFIG_MPEG2_DXVA2_HWACCEL
AV_PIX_FMT_DXVA2_VLD,
--
1.7.9.5

25
packages/multimedia/ffmpeg/patches/ffmpeg-0037-backport-fix-vdpau-vc1-interlace-modes-b37cc5995b88ec68a68cb8e496a008e1cd467077.patch
View File

@ -1,25 +0,0 @@
From b37cc5995b88ec68a68cb8e496a008e1cd467077 Mon Sep 17 00:00:00 2001
From: Rainer Hochecker <fernetmenta@online.de>
Date: Sun, 5 May 2013 15:12:59 +0200
Subject: [PATCH] fix vdpau vc1 interlace modes
---
libavcodec/vdpau_vc1.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/libavcodec/vdpau_vc1.c b/libavcodec/vdpau_vc1.c
index f5da9bb..993ef26 100644
--- a/libavcodec/vdpau_vc1.c
+++ b/libavcodec/vdpau_vc1.c
@@ -59,7 +59,7 @@ static int vdpau_vc1_start_frame(AVCodecContext *avctx,
else
info->picture_type = s->pict_type - 1 + s->pict_type / 3;
- info->frame_coding_mode = v->fcm;
+ info->frame_coding_mode = v->fcm ? v->fcm + 1 : 0;
info->postprocflag = v->postprocflag;
info->pulldown = v->broadcast;
info->interlace = v->interlace;
--
1.7.9.5

236
packages/multimedia/ffmpeg/patches/ffmpeg-0039-h264-expose-h264-frame-packing-as-stereo_mode-metada.patch
View File

@ -1,236 +0,0 @@
From 3a6ea59e122d5bc981a2a5c8f9bafcba0f9f8754 Mon Sep 17 00:00:00 2001
From: Joakim Plate <elupus@ecce.se>
Date: Wed, 26 Jun 2013 22:07:28 +0200
Subject: [PATCH 1/1] h264: expose h264 frame packing as stereo_mode metadata
---
libavcodec/h264.c | 3 ++
libavcodec/h264.h | 35 ++++++++++++++-
libavcodec/h264_parser.c | 1 +
libavcodec/h264_sei.c | 85 +++++++++++++++++++++++++++++++++++++
4 files changed, 123 insertions(+), 1 deletion(-)
diff --git a/libavcodec/h264.c b/libavcodec/h264.c
index 937ad7a..85f2372 100644
--- a/libavcodec/h264.c
+++ b/libavcodec/h264.c
@@ -1437,6 +1437,7 @@ av_cold int ff_h264_decode_init(AVCodecContext *avctx)
h->prev_poc_msb = 1 << 16;
h->prev_frame_num = -1;
h->x264_build = -1;
+ h->sei_fpa.frame_packing_arrangement_cancel_flag = -1;
ff_h264_reset_sei(h);
if (avctx->codec_id == AV_CODEC_ID_H264) {
if (avctx->ticks_per_frame == 1) {
@@ -4783,6 +4784,7 @@ static int decode_frame(AVCodecContext *avctx, void *data,
out->f.reference &= ~DELAYED_PIC_REF;
*got_frame = 1;
*pict = out->f;
+ av_dict_set(&pict->metadata, "stereo_mode", ff_h264_sei_stereo_mode(h), 0);
}
return buf_index;
@@ -4839,6 +4841,7 @@ not_extra:
if (h->next_output_pic && (h->next_output_pic->sync || h->sync>1)) {
*got_frame = 1;
*pict = h->next_output_pic->f;
+ av_dict_set(&pict->metadata, "stereo_mode", ff_h264_sei_stereo_mode(h), 0);
}
}
diff --git a/libavcodec/h264.h b/libavcodec/h264.h
index 7ea613c..15a0b21 100644
--- a/libavcodec/h264.h
+++ b/libavcodec/h264.h
@@ -126,7 +126,8 @@ typedef enum {
SEI_TYPE_PIC_TIMING = 1, ///< picture timing
SEI_TYPE_USER_DATA_ITU_T_T35 = 4, ///< user data registered by ITU-T Recommendation T.35
SEI_TYPE_USER_DATA_UNREGISTERED = 5, ///< unregistered user data
- SEI_TYPE_RECOVERY_POINT = 6 ///< recovery point (frame # to decoder sync)
+ SEI_TYPE_RECOVERY_POINT = 6, ///< recovery point (frame # to decoder sync)
+ SEI_TYPE_FRAME_PACKING = 45, ///< frame packing arrangement
} SEI_Type;
/**
@@ -145,6 +146,19 @@ typedef enum {
} SEI_PicStructType;
/**
+ * frame_packing_arrangement types
+ */
+typedef enum {
+ SEI_FPA_TYPE_CHECKERBOARD = 0,
+ SEI_FPA_TYPE_INTERLEAVE_COLUMN = 1,
+ SEI_FPA_TYPE_INTERLEAVE_ROW = 2,
+ SEI_FPA_TYPE_SIDE_BY_SIDE = 3,
+ SEI_FPA_TYPE_TOP_BOTTOM = 4,
+ SEI_FPA_TYPE_INTERLEAVE_TEMPORAL = 5,
+ SEI_FPA_TYPE_2D = 6,
+} SEI_FpaType;
+
+/**
* Sequence parameter set
*/
typedef struct SPS {
@@ -230,6 +244,17 @@ typedef struct PPS {
} PPS;
/**
+ * Frame Packing Arrangement Type
+ */
+typedef struct FPA {
+ int frame_packing_arrangement_id;
+ int frame_packing_arrangement_cancel_flag; ///< is previous arrangement canceled, -1 if never received
+ SEI_FpaType frame_packing_arrangement_type;
+ int content_interpretation_type;
+ int quincunx_sampling_flag;
+} FPA;
+
+/**
* Memory management control operation opcode.
*/
typedef enum MMCOOpcode {
@@ -626,6 +651,8 @@ typedef struct H264Context {
*/
int valid_recovery_point;
+ FPA sei_fpa;
+
int luma_weight_flag[2]; ///< 7.4.3.2 luma_weight_lX_flag
int chroma_weight_flag[2]; ///< 7.4.3.2 chroma_weight_lX_flag
@@ -771,6 +798,12 @@ void ff_h264_filter_mb(H264Context *h, int mb_x, int mb_y,
*/
void ff_h264_reset_sei(H264Context *h);
+/**
+ * Get stereo_mode string from the h264 frame_packing_arrangement
+ * @param h H.264 context.
+ */
+const char* ff_h264_sei_stereo_mode(H264Context *h);
+
/*
* o-o o-o
* / / /
diff --git a/libavcodec/h264_parser.c b/libavcodec/h264_parser.c
index 44b92b7..aff9ba1 100644
--- a/libavcodec/h264_parser.c
+++ b/libavcodec/h264_parser.c
@@ -158,6 +158,7 @@ static inline int parse_nal_units(AVCodecParserContext *s,
h->sei_dpb_output_delay = 0;
h->sei_cpb_removal_delay = -1;
h->sei_buffering_period_present = 0;
+ h->sei_fpa.frame_packing_arrangement_cancel_flag = -1;
if (!buf_size)
return 0;
diff --git a/libavcodec/h264_sei.c b/libavcodec/h264_sei.c
index ece54f1..fc9b96f 100644
--- a/libavcodec/h264_sei.c
+++ b/libavcodec/h264_sei.c
@@ -194,6 +194,43 @@ static int decode_buffering_period(H264Context *h){
return 0;
}
+static int decode_frame_packing(H264Context *h, int size){
+ int bits = get_bits_left(&h->gb);
+
+ h->sei_fpa.frame_packing_arrangement_id = get_ue_golomb(&h->gb);
+ h->sei_fpa.frame_packing_arrangement_cancel_flag = get_bits(&h->gb, 1);
+ if (!h->sei_fpa.frame_packing_arrangement_cancel_flag) {
+ h->sei_fpa.frame_packing_arrangement_type = get_bits(&h->gb, 7);
+ h->sei_fpa.quincunx_sampling_flag = get_bits(&h->gb, 1);
+ h->sei_fpa.content_interpretation_type = get_bits(&h->gb, 6);
+ skip_bits(&h->gb, 1); /* spatial_flipping_flag */
+ skip_bits(&h->gb, 1); /* frame0_flipped_flag */
+ skip_bits(&h->gb, 1); /* field_views_flag */
+ skip_bits(&h->gb, 1); /* current_frame_is_frame0_flag */
+ skip_bits(&h->gb, 1); /* frame0_self_contained_flag */
+ skip_bits(&h->gb, 1); /* frame1_self_contained_flag */
+ if (!h->sei_fpa.quincunx_sampling_flag && h->sei_fpa.frame_packing_arrangement_type != 5) {
+ skip_bits(&h->gb, 4); /* frame0_grid_position_x */
+ skip_bits(&h->gb, 4); /* frame0_grid_position_y */
+ skip_bits(&h->gb, 4); /* frame1_grid_position_x */
+ skip_bits(&h->gb, 4); /* frame1_grid_position_y */
+ }
+ skip_bits(&h->gb, 8); /* frame_packing_arrangement_reserved_byte */
+ get_ue_golomb(&h->gb) /* frame_packing_arrangement_repetition_period */;
+ }
+ skip_bits(&h->gb, 1); /* frame_packing_arrangement_extension_flag */
+
+ if (h->avctx->debug & FF_DEBUG_PICT_INFO)
+ av_log(h->avctx, AV_LOG_DEBUG, "SEI FPA %d %d %d %d %d\n",
+ h->sei_fpa.frame_packing_arrangement_id,
+ h->sei_fpa.frame_packing_arrangement_cancel_flag,
+ h->sei_fpa.frame_packing_arrangement_type,
+ h->sei_fpa.quincunx_sampling_flag,
+ h->sei_fpa.content_interpretation_type);
+ skip_bits(&h->gb, 8*size - (bits - get_bits_left(&h->gb)));
+ return 0;
+}
+
int ff_h264_decode_sei(H264Context *h){
while (get_bits_left(&h->gb) > 16) {
int size, type;
@@ -236,6 +273,9 @@ int ff_h264_decode_sei(H264Context *h){
if(decode_buffering_period(h) < 0)
return -1;
break;
+ case SEI_TYPE_FRAME_PACKING:
+ if(decode_frame_packing(h, size) < 0)
+ return -1;
default:
skip_bits(&h->gb, 8*size);
}
@@ -246,3 +286,48 @@ int ff_h264_decode_sei(H264Context *h){
return 0;
}
+
+const char* ff_h264_sei_stereo_mode(H264Context *h)
+{
+ if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 0) {
+ switch (h->sei_fpa.frame_packing_arrangement_type) {
+ case SEI_FPA_TYPE_CHECKERBOARD:
+ if (h->sei_fpa.content_interpretation_type == 2)
+ return "checkerboard_rl";
+ else
+ return "checkerboard_lr";
+ case SEI_FPA_TYPE_INTERLEAVE_COLUMN:
+ if (h->sei_fpa.content_interpretation_type == 2)
+ return "col_interleaved_rl";
+ else
+ return "col_interleaved_lr";
+ case SEI_FPA_TYPE_INTERLEAVE_ROW:
+ if (h->sei_fpa.content_interpretation_type == 2)
+ return "row_interleaved_rl";
+ else
+ return "row_interleaved_lr";
+ case SEI_FPA_TYPE_SIDE_BY_SIDE:
+ if (h->sei_fpa.content_interpretation_type == 2)
+ return "right_left";
+ else
+ return "left_right";
+ case SEI_FPA_TYPE_TOP_BOTTOM:
+ if (h->sei_fpa.content_interpretation_type == 2)
+ return "bottom_top";
+ else
+ return "top_bottom";
+ case SEI_FPA_TYPE_INTERLEAVE_TEMPORAL:
+ if (h->sei_fpa.content_interpretation_type == 2)
+ return "block_rl";
+ else
+ return "block_lr";
+ case SEI_FPA_TYPE_2D:
+ default:
+ return "mono";
+ }
+ } else if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 1) {
+ return "mono";
+ } else {
+ return NULL;
+ }
+}
--
1.8.2

311
packages/multimedia/ffmpeg/patches/ffmpeg-0040-ffmpeg-backport-arm-Add-VFP-accelerated-version-of-s.patch
View File

@ -1,311 +0,0 @@
From 40daea3c1bafa9cea37b65f856c3c0432767d760 Mon Sep 17 00:00:00 2001
From: Ben Avison <bavison@riscosopen.org>
Date: Mon, 15 Jul 2013 18:28:09 +0100
Subject: [PATCH 39/49] [ffmpeg] - backport - arm: Add VFP-accelerated version
of synth_filter_float
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Before After
Mean StdDev Mean StdDev Change
This function 9295.0 114.9 4853.2 83.5 +91.5%
Overall 23699.8 397.6 19285.5 292.0 +22.9%
Signed-off-by: Martin Storsjö <martin@martin.st>
---
libavcodec/arm/Makefile | 1 +
libavcodec/arm/fft_init_arm.c | 8 +
libavcodec/arm/synth_filter_vfp.S | 243 ++++++++++++++++++++++++++
3 files changed, 252 insertions(+)
create mode 100644 libavcodec/arm/synth_filter_vfp.S
diff --git a/libavcodec/arm/Makefile b/libavcodec/arm/Makefile
index 1c91d62..aee9d73 100644
--- a/libavcodec/arm/Makefile
+++ b/libavcodec/arm/Makefile
@@ -58,6 +58,7 @@ ARMV6-OBJS += arm/dsputil_init_armv6.o \
arm/dsputil_armv6.o \
arm/simple_idct_armv6.o \
+VFP-OBJS-$(CONFIG_DCA_DECODER) += arm/synth_filter_vfp.o
VFP-OBJS-$(HAVE_ARMV6) += arm/fmtconvert_vfp.o
NEON-OBJS-$(CONFIG_FFT) += arm/fft_neon.o \
diff --git a/libavcodec/arm/fft_init_arm.c b/libavcodec/arm/fft_init_arm.c
index 8c98abc..fe0acc5 100644
--- a/libavcodec/arm/fft_init_arm.c
+++ b/libavcodec/arm/fft_init_arm.c
@@ -32,6 +32,12 @@ void ff_mdct_calc_neon(FFTContext *s, FFTSample *output, const FFTSample *input)
void ff_rdft_calc_neon(struct RDFTContext *s, FFTSample *z);
+void ff_synth_filter_float_vfp(FFTContext *imdct,
+ float *synth_buf_ptr, int *synth_buf_offset,
+ float synth_buf2[32], const float window[512],
+ float out[32], const float in[32],
+ float scale);
+
void ff_synth_filter_float_neon(FFTContext *imdct,
float *synth_buf_ptr, int *synth_buf_offset,
float synth_buf2[32], const float window[512],
@@ -71,6 +77,8 @@ av_cold void ff_synth_filter_init_arm(SynthFilterContext *s)
{
int cpu_flags = av_get_cpu_flags();
+ if (have_vfp(cpu_flags) && !have_vfpv3(cpu_flags))
+ s->synth_filter_float = ff_synth_filter_float_vfp;
if (have_neon(cpu_flags))
s->synth_filter_float = ff_synth_filter_float_neon;
}
diff --git a/libavcodec/arm/synth_filter_vfp.S b/libavcodec/arm/synth_filter_vfp.S
new file mode 100644
index 0000000..c219c41
--- /dev/null
+++ b/libavcodec/arm/synth_filter_vfp.S
@@ -0,0 +1,243 @@
+/*
+ * Copyright (c) 2013 RISC OS Open Ltd
+ * Author: Ben Avison <bavison@riscosopen.org>
+ *
+ * This file is part of Libav.
+ *
+ * Libav 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.
+ *
+ * Libav 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 Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/arm/asm.S"
+
+IMDCT .req r0
+ORIG_P_SB .req r1
+P_SB_OFF .req r2
+I .req r0
+P_SB2_UP .req r1
+OLDFPSCR .req r2
+P_SB2_DN .req r3
+P_WIN_DN .req r4
+P_OUT_DN .req r5
+P_SB .req r6
+J_WRAP .req r7
+P_WIN_UP .req r12
+P_OUT_UP .req r14
+
+SCALE .req s0
+SBUF_DAT_REV0 .req s4
+SBUF_DAT_REV1 .req s5
+SBUF_DAT_REV2 .req s6
+SBUF_DAT_REV3 .req s7
+VA0 .req s8
+VA3 .req s11
+VB0 .req s12
+VB3 .req s15
+VC0 .req s8
+VC3 .req s11
+VD0 .req s12
+VD3 .req s15
+SBUF_DAT0 .req s16
+SBUF_DAT1 .req s17
+SBUF_DAT2 .req s18
+SBUF_DAT3 .req s19
+SBUF_DAT_ALT0 .req s20
+SBUF_DAT_ALT1 .req s21
+SBUF_DAT_ALT2 .req s22
+SBUF_DAT_ALT3 .req s23
+WIN_DN_DAT0 .req s24
+WIN_UP_DAT0 .req s28
+
+
+.macro inner_loop half, tail, head
+ .if (OFFSET & (64*4)) == 0 @ even numbered call
+ SBUF_DAT_THIS0 .req SBUF_DAT0
+ SBUF_DAT_THIS1 .req SBUF_DAT1
+ SBUF_DAT_THIS2 .req SBUF_DAT2
+ SBUF_DAT_THIS3 .req SBUF_DAT3
+ .ifnc "\head",""
+ vldr d8, [P_SB, #OFFSET] @ d8 = SBUF_DAT
+ vldr d9, [P_SB, #OFFSET+8]
+ .endif
+ .else
+ SBUF_DAT_THIS0 .req SBUF_DAT_ALT0
+ SBUF_DAT_THIS1 .req SBUF_DAT_ALT1
+ SBUF_DAT_THIS2 .req SBUF_DAT_ALT2
+ SBUF_DAT_THIS3 .req SBUF_DAT_ALT3
+ .ifnc "\head",""
+ vldr d10, [P_SB, #OFFSET] @ d10 = SBUF_DAT_ALT
+ vldr d11, [P_SB, #OFFSET+8]
+ .endif
+ .endif
+ .ifnc "\tail",""
+ .ifc "\half","ab"
+ vmls.f VA0, SBUF_DAT_REV0, WIN_DN_DAT0 @ all operands treated as vectors
+ .else
+ vmla.f VD0, SBUF_DAT_REV0, WIN_DN_DAT0 @ all operands treated as vectors
+ .endif
+ .endif
+ .ifnc "\head",""
+ vldr d14, [P_WIN_UP, #OFFSET] @ d14 = WIN_UP_DAT
+ vldr d15, [P_WIN_UP, #OFFSET+8]
+ vldr d12, [P_WIN_DN, #OFFSET] @ d12 = WIN_DN_DAT
+ vldr d13, [P_WIN_DN, #OFFSET+8]
+ vmov SBUF_DAT_REV3, SBUF_DAT_THIS0
+ vmov SBUF_DAT_REV2, SBUF_DAT_THIS1
+ vmov SBUF_DAT_REV1, SBUF_DAT_THIS2
+ vmov SBUF_DAT_REV0, SBUF_DAT_THIS3
+ .ifc "\half","ab"
+ vmla.f VB0, SBUF_DAT_THIS0, WIN_UP_DAT0
+ .else
+ vmla.f VC0, SBUF_DAT_THIS0, WIN_UP_DAT0
+ .endif
+ teq J_WRAP, #J
+ bne 2f @ strongly predictable, so better than cond exec in this case
+ sub P_SB, P_SB, #512*4
+2:
+ .set J, J - 64
+ .set OFFSET, OFFSET + 64*4
+ .endif
+ .unreq SBUF_DAT_THIS0
+ .unreq SBUF_DAT_THIS1
+ .unreq SBUF_DAT_THIS2
+ .unreq SBUF_DAT_THIS3
+.endm
+
+
+/* void ff_synth_filter_float_vfp(FFTContext *imdct,
+ * float *synth_buf_ptr, int *synth_buf_offset,
+ * float synth_buf2[32], const float window[512],
+ * float out[32], const float in[32], float scale)
+ */
+function ff_synth_filter_float_vfp, export=1
+ push {r3-r7,lr}
+ vpush {s16-s31}
+ ldr lr, [P_SB_OFF]
+ add a2, ORIG_P_SB, lr, LSL #2 @ calculate synth_buf to pass to imdct_half
+ mov P_SB, a2 @ and keep a copy for ourselves
+ bic J_WRAP, lr, #63 @ mangled to make testing for wrap easier in inner loop
+ sub lr, lr, #32
+ and lr, lr, #512-32
+ str lr, [P_SB_OFF] @ rotate offset, modulo buffer size, ready for next call
+ ldr a3, [sp, #(16+6+2)*4] @ fetch in from stack, to pass to imdct_half
+VFP vmov s16, SCALE @ imdct_half is free to corrupt s0, but it contains one of our arguments in hardfp case
+ bl ff_imdct_half_vfp
+VFP vmov SCALE, s16
+
+ fmrx OLDFPSCR, FPSCR
+ ldr lr, =0x03030000 @ RunFast mode, short vectors of length 4, stride 1
+ fmxr FPSCR, lr
+ ldr P_SB2_DN, [sp, #16*4]
+ ldr P_WIN_DN, [sp, #(16+6+0)*4]
+ ldr P_OUT_DN, [sp, #(16+6+1)*4]
+NOVFP vldr SCALE, [sp, #(16+6+3)*4]
+
+#define IMM_OFF_SKEW 956 /* also valid immediate constant when you add 16*4 */
+ add P_SB, P_SB, #IMM_OFF_SKEW @ so we can use -ve offsets to use full immediate offset range
+ add P_SB2_UP, P_SB2_DN, #16*4
+ add P_WIN_UP, P_WIN_DN, #16*4+IMM_OFF_SKEW
+ add P_OUT_UP, P_OUT_DN, #16*4
+ add P_SB2_DN, P_SB2_DN, #16*4
+ add P_WIN_DN, P_WIN_DN, #12*4+IMM_OFF_SKEW
+ add P_OUT_DN, P_OUT_DN, #16*4
+ mov I, #4
+1:
+ vldmia P_SB2_UP!, {VB0-VB3}
+ vldmdb P_SB2_DN!, {VA0-VA3}
+ .set J, 512 - 64
+ .set OFFSET, -IMM_OFF_SKEW
+ inner_loop ab,, head
+ .rept 7
+ inner_loop ab, tail, head
+ .endr
+ inner_loop ab, tail
+ add P_WIN_UP, P_WIN_UP, #4*4
+ sub P_WIN_DN, P_WIN_DN, #4*4
+ vmul.f VB0, VB0, SCALE @ SCALE treated as scalar
+ add P_SB, P_SB, #(512+4)*4
+ subs I, I, #1
+ vmul.f VA0, VA0, SCALE
+ vstmia P_OUT_UP!, {VB0-VB3}
+ vstmdb P_OUT_DN!, {VA0-VA3}
+ bne 1b
+
+ add P_SB2_DN, P_SB2_DN, #(16+28-12)*4
+ sub P_SB2_UP, P_SB2_UP, #(16+16)*4
+ add P_WIN_DN, P_WIN_DN, #(32+16+28-12)*4
+ mov I, #4
+1:
+ vldr.d d4, zero @ d4 = VC0
+ vldr.d d5, zero
+ vldr.d d6, zero @ d6 = VD0
+ vldr.d d7, zero
+ .set J, 512 - 64
+ .set OFFSET, -IMM_OFF_SKEW
+ inner_loop cd,, head
+ .rept 7
+ inner_loop cd, tail, head
+ .endr
+ inner_loop cd, tail
+ add P_WIN_UP, P_WIN_UP, #4*4
+ sub P_WIN_DN, P_WIN_DN, #4*4
+ add P_SB, P_SB, #(512+4)*4
+ subs I, I, #1
+ vstmia P_SB2_UP!, {VC0-VC3}
+ vstmdb P_SB2_DN!, {VD0-VD3}
+ bne 1b
+
+ fmxr FPSCR, OLDFPSCR
+ vpop {s16-s31}
+ pop {r3-r7,pc}
+endfunc
+
+ .unreq IMDCT
+ .unreq ORIG_P_SB
+ .unreq P_SB_OFF
+ .unreq I
+ .unreq P_SB2_UP
+ .unreq OLDFPSCR
+ .unreq P_SB2_DN
+ .unreq P_WIN_DN
+ .unreq P_OUT_DN
+ .unreq P_SB
+ .unreq J_WRAP
+ .unreq P_WIN_UP
+ .unreq P_OUT_UP
+
+ .unreq SCALE
+ .unreq SBUF_DAT_REV0
+ .unreq SBUF_DAT_REV1
+ .unreq SBUF_DAT_REV2
+ .unreq SBUF_DAT_REV3
+ .unreq VA0
+ .unreq VA3
+ .unreq VB0
+ .unreq VB3
+ .unreq VC0
+ .unreq VC3
+ .unreq VD0
+ .unreq VD3
+ .unreq SBUF_DAT0
+ .unreq SBUF_DAT1
+ .unreq SBUF_DAT2
+ .unreq SBUF_DAT3
+ .unreq SBUF_DAT_ALT0
+ .unreq SBUF_DAT_ALT1
+ .unreq SBUF_DAT_ALT2
+ .unreq SBUF_DAT_ALT3
+ .unreq WIN_DN_DAT0
+ .unreq WIN_UP_DAT0
+
+ .align 3
+zero: .word 0, 0
--
1.7.9.5

102
packages/multimedia/ffmpeg/patches/ffmpeg-0041-ffmpeg-backport-arm-Add-VFP-accelerated-version-of-i.patch
View File

@ -1,102 +0,0 @@
From 8ead63b22d31bf71976fc6964922b43d8e0d660b Mon Sep 17 00:00:00 2001
From: Ben Avison <bavison@riscosopen.org>
Date: Mon, 15 Jul 2013 18:28:10 +0100
Subject: [PATCH 40/49] [ffmpeg] - backport - arm: Add VFP-accelerated version
of int32_to_float_fmul_scalar
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Before After
Mean StdDev Mean StdDev Change
This function 1175.0 4.4 366.2 18.3 +220.8%
Overall 19285.5 292.0 18420.5 489.1 +4.7%
Signed-off-by: Martin Storsjö <martin@martin.st>
---
libavcodec/arm/fmtconvert_init_arm.c | 10 ++++++
libavcodec/arm/fmtconvert_vfp.S | 38 +++++++++++++++++++++++
2 files changed, 48 insertions(+)
diff --git a/libavcodec/arm/fmtconvert_init_arm.c b/libavcodec/arm/fmtconvert_init_arm.c
index 1d99c97..de3b78b 100644
--- a/libavcodec/arm/fmtconvert_init_arm.c
+++ b/libavcodec/arm/fmtconvert_init_arm.c
@@ -28,6 +28,9 @@
void ff_int32_to_float_fmul_scalar_neon(float *dst, const int *src,
float mul, int len);
+void ff_int32_to_float_fmul_scalar_vfp(float *dst, const int32_t *src,
+ float mul, int len);
+
void ff_float_to_int16_neon(int16_t *dst, const float *src, long len);
void ff_float_to_int16_interleave_neon(int16_t *, const float **, long, int);
@@ -38,6 +41,13 @@ av_cold void ff_fmt_convert_init_arm(FmtConvertContext *c, AVCodecContext *avctx
int cpu_flags = av_get_cpu_flags();
if (have_vfp(cpu_flags) && have_armv6(cpu_flags)) {
+ if (!have_vfpv3(cpu_flags)) {
+ // This function doesn't use anything armv6 specific in itself,
+ // but ff_float_to_int16_vfp which is in the same assembly source
+ // file does, thus the whole file requires armv6 to be built.
+ c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_vfp;
+ }
+
c->float_to_int16 = ff_float_to_int16_vfp;
}
diff --git a/libavcodec/arm/fmtconvert_vfp.S b/libavcodec/arm/fmtconvert_vfp.S
index 7b012bc..3cc3e56 100644
--- a/libavcodec/arm/fmtconvert_vfp.S
+++ b/libavcodec/arm/fmtconvert_vfp.S
@@ -1,5 +1,6 @@
/*
* Copyright (c) 2008 Siarhei Siamashka <ssvb@users.sourceforge.net>
+ * Copyright (c) 2013 RISC OS Open Ltd <bavison@riscosopen.org>
*
* This file is part of FFmpeg.
*
@@ -76,3 +77,40 @@ function ff_float_to_int16_vfp, export=1
vpop {d8-d11}
pop {r4-r8,pc}
endfunc
+
+/**
+ * ARM VFP optimised int32 to float conversion.
+ * Assume len is a multiple of 8, destination buffer is at least 4 bytes aligned
+ * (16 bytes alignment is best for BCM2835), little-endian.
+ */
+@ void ff_int32_to_float_fmul_scalar_vfp(float *dst, const int32_t *src, float mul, int len)
+function ff_int32_to_float_fmul_scalar_vfp, export=1
+VFP tmp .req a4
+VFP len .req a3
+NOVFP tmp .req a3
+NOVFP len .req a4
+NOVFP vmov s0, a3
+ ldr tmp, =0x03070000 @ RunFast mode, short vectors of length 8, stride 1
+ fmrx ip, FPSCR
+ fmxr FPSCR, tmp
+1:
+ vldmia a2!, {s8-s15}
+ vcvt.f32.s32 s8, s8
+ vcvt.f32.s32 s9, s9
+ vcvt.f32.s32 s10, s10
+ vcvt.f32.s32 s11, s11
+ vcvt.f32.s32 s12, s12
+ vcvt.f32.s32 s13, s13
+ vcvt.f32.s32 s14, s14
+ vcvt.f32.s32 s15, s15
+ vmul.f32 s8, s8, s0
+ subs len, len, #8
+ vstmia a1!, {s8-s11}
+ vstmia a1!, {s12-s15}
+ bne 1b
+
+ fmxr FPSCR, ip
+ bx lr
+endfunc
+ .unreq tmp
+ .unreq len
--
1.7.9.5

78
packages/multimedia/ffmpeg/patches/ffmpeg-0042-ffmpeg-backport-fmtconvert-Add-a-new-method-int32_to.patch
View File

@ -1,78 +0,0 @@
From 7901e7216cf6406a2ea430c71af94ebee72f262b Mon Sep 17 00:00:00 2001
From: Ben Avison <bavison@riscosopen.org>
Date: Mon, 15 Jul 2013 18:28:11 +0100
Subject: [PATCH 41/49] [ffmpeg] - backport - fmtconvert: Add a new method,
int32_to_float_fmul_array8
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
This is similar to int32_to_float_fmul_scalar, but
loads a new scalar multiplier every 8 input samples.
This enables the use of much larger input arrays, which
is important for pipelining on some CPUs (such as
ARMv6).
Signed-off-by: Martin Storsjö <martin@martin.st>
---
libavcodec/fmtconvert.c | 10 ++++++++++
libavcodec/fmtconvert.h | 16 ++++++++++++++++
2 files changed, 26 insertions(+)
diff --git a/libavcodec/fmtconvert.c b/libavcodec/fmtconvert.c
index 79e9645..1c45d35 100644
--- a/libavcodec/fmtconvert.c
+++ b/libavcodec/fmtconvert.c
@@ -30,6 +30,15 @@ static void int32_to_float_fmul_scalar_c(float *dst, const int *src, float mul,
dst[i] = src[i] * mul;
}
+static void int32_to_float_fmul_array8_c(FmtConvertContext *c, float *dst,
+ const int32_t *src, const float *mul,
+ int len)
+{
+ int i;
+ for (i = 0; i < len; i += 8)
+ c->int32_to_float_fmul_scalar(&dst[i], &src[i], *mul++, 8);
+}
+
static av_always_inline int float_to_int16_one(const float *src){
return av_clip_int16(lrintf(*src));
}
@@ -79,6 +88,7 @@ void ff_float_interleave_c(float *dst, const float **src, unsigned int len,
av_cold void ff_fmt_convert_init(FmtConvertContext *c, AVCodecContext *avctx)
{
c->int32_to_float_fmul_scalar = int32_to_float_fmul_scalar_c;
+ c->int32_to_float_fmul_array8 = int32_to_float_fmul_array8_c;
c->float_to_int16 = float_to_int16_c;
c->float_to_int16_interleave = float_to_int16_interleave_c;
c->float_interleave = ff_float_interleave_c;
diff --git a/libavcodec/fmtconvert.h b/libavcodec/fmtconvert.h
index 3fb9f4e..02468dc 100644
--- a/libavcodec/fmtconvert.h
+++ b/libavcodec/fmtconvert.h
@@ -38,6 +38,22 @@ typedef struct FmtConvertContext {
void (*int32_to_float_fmul_scalar)(float *dst, const int *src, float mul, int len);
/**
+ * Convert an array of int32_t to float and multiply by a float value from another array,
+ * stepping along the float array once for each 8 integers.
+ * @param c pointer to FmtConvertContext.
+ * @param dst destination array of float.
+ * constraints: 16-byte aligned
+ * @param src source array of int32_t.
+ * constraints: 16-byte aligned
+ * @param mul source array of float multipliers.
+ * @param len number of elements to convert.
+ * constraints: multiple of 8
+ */
+ void (*int32_to_float_fmul_array8)(struct FmtConvertContext *c,
+ float *dst, const int32_t *src,
+ const float *mul, int len);
+
+ /**
* Convert an array of float to an array of int16_t.
*
* Convert floats from in the range [-32768.0,32767.0] to ints
--
1.7.9.5

90
packages/multimedia/ffmpeg/patches/ffmpeg-0043-ffmpeg-backport-dcadec-Use-int32_to_float_fmul_array.patch
View File

@ -1,90 +0,0 @@
From fa755fe82fe4cfbb85b7c57501912da2e1f316bc Mon Sep 17 00:00:00 2001
From: Ben Avison <bavison@riscosopen.org>
Date: Tue, 16 Jul 2013 15:41:18 +0300
Subject: [PATCH 42/49] [ffmpeg] - backport - dcadec: Use
int32_to_float_fmul_array8
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Signed-off-by: Martin Storsjö <martin@martin.st>
---
libavcodec/dcadec.c | 23 +++++++++++++++--------
1 file changed, 15 insertions(+), 8 deletions(-)
diff --git a/libavcodec/dcadec.c b/libavcodec/dcadec.c
index 1b955e4..b648613 100644
--- a/libavcodec/dcadec.c
+++ b/libavcodec/dcadec.c
@@ -1302,7 +1302,7 @@ static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
/* FIXME */
float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
- LOCAL_ALIGNED_16(int, block, [8]);
+ LOCAL_ALIGNED_16(int, block, [8 * DCA_SUBBANDS]);
/*
* Audio data
@@ -1315,6 +1315,8 @@ static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
quant_step_table = lossy_quant_d;
for (k = base_channel; k < s->prim_channels; k++) {
+ float rscale[DCA_SUBBANDS];
+
if (get_bits_left(&s->gb) < 0)
return AVERROR_INVALIDDATA;
@@ -1337,11 +1339,12 @@ static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
* Extract bits from the bit stream
*/
if (!abits) {
- memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
+ rscale[l] = 0;
+ memset(block + 8 * l, 0, 8 * sizeof(block[0]));
} else {
/* Deal with transients */
int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
- float rscale = quant_step_size * s->scale_factor[k][l][sfi] *
+ rscale[l] = quant_step_size * s->scale_factor[k][l][sfi] *
s->scalefactor_adj[k][sel];
if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
@@ -1355,7 +1358,7 @@ static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
block_code1 = get_bits(&s->gb, size);
block_code2 = get_bits(&s->gb, size);
err = decode_blockcodes(block_code1, block_code2,
- levels, block);
+ levels, block + 8 * l);
if (err) {
av_log(s->avctx, AV_LOG_ERROR,
"ERROR: block code look-up failed\n");
@@ -1364,19 +1367,23 @@ static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
} else {
/* no coding */
for (m = 0; m < 8; m++)
- block[m] = get_sbits(&s->gb, abits - 3);
+ block[8 * l + m] = get_sbits(&s->gb, abits - 3);
}
} else {
/* Huffman coded */
for (m = 0; m < 8; m++)
- block[m] = get_bitalloc(&s->gb,
+ block[8 * l + m] = get_bitalloc(&s->gb,
&dca_smpl_bitalloc[abits], sel);
}
- s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
- block, rscale, 8);
}
+ }
+ s->fmt_conv.int32_to_float_fmul_array8(&s->fmt_conv, subband_samples[k][0],
+ block, rscale, 8 * s->vq_start_subband[k]);
+
+ for (l = 0; l < s->vq_start_subband[k]; l++) {
+ int m;
/*
* Inverse ADPCM if in prediction mode
*/
--
1.7.9.5

222
packages/multimedia/ffmpeg/patches/ffmpeg-0044-ffmpeg-backport-arm-Add-VFP-accelerated-version-of-i.patch
View File

@ -1,222 +0,0 @@
From c908a710261f33130569c4360175d8f19a282d67 Mon Sep 17 00:00:00 2001
From: Ben Avison <bavison@riscosopen.org>
Date: Mon, 15 Jul 2013 18:28:12 +0100
Subject: [PATCH 43/49] [ffmpeg] - backport - arm: Add VFP-accelerated version
of int32_to_float_fmul_array8
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Before After
Mean StdDev Mean StdDev Change
This function 366.2 18.3 277.8 13.7 +31.9%
Overall 18420.5 489.1 17049.5 408.2 +8.0%
Signed-off-by: Martin Storsjö <martin@martin.st>
---
libavcodec/arm/fmtconvert_init_arm.c | 6 +-
libavcodec/arm/fmtconvert_vfp.S | 162 +++++++++++++++++++++++
2 files changed, 167 insertions(+), 1 deletion(-)
diff --git a/libavcodec/arm/fmtconvert_init_arm.c b/libavcodec/arm/fmtconvert_init_arm.c
index de3b78b..92d94a0 100644
--- a/libavcodec/arm/fmtconvert_init_arm.c
+++ b/libavcodec/arm/fmtconvert_init_arm.c
@@ -30,6 +30,9 @@ void ff_int32_to_float_fmul_scalar_neon(float *dst, const int *src,
void ff_int32_to_float_fmul_scalar_vfp(float *dst, const int32_t *src,
float mul, int len);
+void ff_int32_to_float_fmul_array8_vfp(FmtConvertContext *c, float *dst,
+ const int32_t *src, const float *mul,
+ int len);
void ff_float_to_int16_neon(int16_t *dst, const float *src, long len);
void ff_float_to_int16_interleave_neon(int16_t *, const float **, long, int);
@@ -42,10 +45,11 @@ av_cold void ff_fmt_convert_init_arm(FmtConvertContext *c, AVCodecContext *avctx
if (have_vfp(cpu_flags) && have_armv6(cpu_flags)) {
if (!have_vfpv3(cpu_flags)) {
- // This function doesn't use anything armv6 specific in itself,
+ // These functions don't use anything armv6 specific in themselves,
// but ff_float_to_int16_vfp which is in the same assembly source
// file does, thus the whole file requires armv6 to be built.
c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_vfp;
+ c->int32_to_float_fmul_array8 = ff_int32_to_float_fmul_array8_vfp;
}
c->float_to_int16 = ff_float_to_int16_vfp;
diff --git a/libavcodec/arm/fmtconvert_vfp.S b/libavcodec/arm/fmtconvert_vfp.S
index 3cc3e56..a6d4ebd 100644
--- a/libavcodec/arm/fmtconvert_vfp.S
+++ b/libavcodec/arm/fmtconvert_vfp.S
@@ -83,6 +83,168 @@ endfunc
* Assume len is a multiple of 8, destination buffer is at least 4 bytes aligned
* (16 bytes alignment is best for BCM2835), little-endian.
*/
+@ void ff_int32_to_float_fmul_array8_vfp(FmtConvertContext *c, float *dst, const int32_t *src, const float *mul, int len)
+function ff_int32_to_float_fmul_array8_vfp, export=1
+ push {lr}
+ ldr a1, [sp, #4]
+ subs lr, a1, #3*8
+ bcc 50f @ too short to pipeline
+ @ Now need to find (len / 8) % 3. The approximation
+ @ x / 24 = (x * 0xAB) >> 12
+ @ is good for x < 4096, which is true for both AC3 and DCA.
+ mov a1, #0xAB
+ ldr ip, =0x03070000 @ RunFast mode, short vectors of length 8, stride 1
+ mul a1, lr, a1
+ vpush {s16-s31}
+ mov a1, a1, lsr #12
+ add a1, a1, a1, lsl #1
+ rsb a1, a1, lr, lsr #3
+ cmp a1, #1
+ fmrx a1, FPSCR
+ fmxr FPSCR, ip
+ beq 11f
+ blo 10f
+ @ Array is (2 + multiple of 3) x 8 floats long
+ @ drop through...
+ vldmia a3!, {s16-s23}
+ vldmia a4!, {s2,s3}
+ vldmia a3!, {s24-s31}
+ vcvt.f32.s32 s16, s16
+ vcvt.f32.s32 s17, s17
+ vcvt.f32.s32 s18, s18
+ vcvt.f32.s32 s19, s19
+ vcvt.f32.s32 s20, s20
+ vcvt.f32.s32 s21, s21
+ vcvt.f32.s32 s22, s22
+ vcvt.f32.s32 s23, s23
+ vmul.f32 s16, s16, s2
+ @ drop through...
+3:
+ vldmia a3!, {s8-s15}
+ vldmia a4!, {s1}
+ vcvt.f32.s32 s24, s24
+ vcvt.f32.s32 s25, s25
+ vcvt.f32.s32 s26, s26
+ vcvt.f32.s32 s27, s27
+ vcvt.f32.s32 s28, s28
+ vcvt.f32.s32 s29, s29
+ vcvt.f32.s32 s30, s30
+ vcvt.f32.s32 s31, s31
+ vmul.f32 s24, s24, s3
+ vstmia a2!, {s16-s19}
+ vstmia a2!, {s20-s23}
+2:
+ vldmia a3!, {s16-s23}
+ vldmia a4!, {s2}
+ vcvt.f32.s32 s8, s8
+ vcvt.f32.s32 s9, s9
+ vcvt.f32.s32 s10, s10
+ vcvt.f32.s32 s11, s11
+ vcvt.f32.s32 s12, s12
+ vcvt.f32.s32 s13, s13
+ vcvt.f32.s32 s14, s14
+ vcvt.f32.s32 s15, s15
+ vmul.f32 s8, s8, s1
+ vstmia a2!, {s24-s27}
+ vstmia a2!, {s28-s31}
+1:
+ vldmia a3!, {s24-s31}
+ vldmia a4!, {s3}
+ vcvt.f32.s32 s16, s16
+ vcvt.f32.s32 s17, s17
+ vcvt.f32.s32 s18, s18
+ vcvt.f32.s32 s19, s19
+ vcvt.f32.s32 s20, s20
+ vcvt.f32.s32 s21, s21
+ vcvt.f32.s32 s22, s22
+ vcvt.f32.s32 s23, s23
+ vmul.f32 s16, s16, s2
+ vstmia a2!, {s8-s11}
+ vstmia a2!, {s12-s15}
+
+ subs lr, lr, #8*3
+ bpl 3b
+
+ vcvt.f32.s32 s24, s24
+ vcvt.f32.s32 s25, s25
+ vcvt.f32.s32 s26, s26
+ vcvt.f32.s32 s27, s27
+ vcvt.f32.s32 s28, s28
+ vcvt.f32.s32 s29, s29
+ vcvt.f32.s32 s30, s30
+ vcvt.f32.s32 s31, s31
+ vmul.f32 s24, s24, s3
+ vstmia a2!, {s16-s19}
+ vstmia a2!, {s20-s23}
+ vstmia a2!, {s24-s27}
+ vstmia a2!, {s28-s31}
+
+ fmxr FPSCR, a1
+ vpop {s16-s31}
+ pop {pc}
+
+10: @ Array is (multiple of 3) x 8 floats long
+ vldmia a3!, {s8-s15}
+ vldmia a4!, {s1,s2}
+ vldmia a3!, {s16-s23}
+ vcvt.f32.s32 s8, s8
+ vcvt.f32.s32 s9, s9
+ vcvt.f32.s32 s10, s10
+ vcvt.f32.s32 s11, s11
+ vcvt.f32.s32 s12, s12
+ vcvt.f32.s32 s13, s13
+ vcvt.f32.s32 s14, s14
+ vcvt.f32.s32 s15, s15
+ vmul.f32 s8, s8, s1
+ b 1b
+
+11: @ Array is (1 + multiple of 3) x 8 floats long
+ vldmia a3!, {s24-s31}
+ vldmia a4!, {s3}
+ vldmia a3!, {s8-s15}
+ vldmia a4!, {s1}
+ vcvt.f32.s32 s24, s24
+ vcvt.f32.s32 s25, s25
+ vcvt.f32.s32 s26, s26
+ vcvt.f32.s32 s27, s27
+ vcvt.f32.s32 s28, s28
+ vcvt.f32.s32 s29, s29
+ vcvt.f32.s32 s30, s30
+ vcvt.f32.s32 s31, s31
+ vmul.f32 s24, s24, s3
+ b 2b
+
+50:
+ ldr lr, =0x03070000 @ RunFast mode, short vectors of length 8, stride 1
+ fmrx ip, FPSCR
+ fmxr FPSCR, lr
+51:
+ vldmia a3!, {s8-s15}
+ vldmia a4!, {s0}
+ vcvt.f32.s32 s8, s8
+ vcvt.f32.s32 s9, s9
+ vcvt.f32.s32 s10, s10
+ vcvt.f32.s32 s11, s11
+ vcvt.f32.s32 s12, s12
+ vcvt.f32.s32 s13, s13
+ vcvt.f32.s32 s14, s14
+ vcvt.f32.s32 s15, s15
+ vmul.f32 s8, s8, s0
+ subs a1, a1, #8
+ vstmia a2!, {s8-s11}
+ vstmia a2!, {s12-s15}
+ bne 51b
+
+ fmxr FPSCR, ip
+ pop {pc}
+endfunc
+
+/**
+ * ARM VFP optimised int32 to float conversion.
+ * Assume len is a multiple of 8, destination buffer is at least 4 bytes aligned
+ * (16 bytes alignment is best for BCM2835), little-endian.
+ * TODO: could be further optimised by unrolling and interleaving, as above
+ */
@ void ff_int32_to_float_fmul_scalar_vfp(float *dst, const int32_t *src, float mul, int len)
function ff_int32_to_float_fmul_scalar_vfp, export=1
VFP tmp .req a4
--
1.7.9.5

274
packages/multimedia/ffmpeg/patches/ffmpeg-0045-ffmpeg-backport-arm-Add-VFP-accelerated-version-of-i.patch
View File

@ -1,274 +0,0 @@
From 15520de67fc951213ab32661b8b368a9439e8b9a Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Martin=20Storsj=C3=B6?= <martin@martin.st>
Date: Fri, 19 Jul 2013 10:59:17 +0300
Subject: [PATCH 44/49] [ffmpeg] - backport - arm: Add VFP-accelerated version
of imdct_half
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Before After
Mean StdDev Mean StdDev Change
This function 2653.0 28.5 1108.8 51.4 +139.3%
Overall 17049.5 408.2 15973.0 223.2 +6.7%
Signed-off-by: Martin Storsjö <martin@martin.st>
---
libavcodec/arm/Makefile | 1 +
libavcodec/arm/fft_init_arm.c | 9 ++
libavcodec/arm/mdct_vfp.S | 205 ++++++++++++++++++++++++++++++
3 files changed, 215 insertions(+)
create mode 100644 libavcodec/arm/mdct_vfp.S
diff --git a/libavcodec/arm/Makefile b/libavcodec/arm/Makefile
index aee9d73..27e80d5 100644
--- a/libavcodec/arm/Makefile
+++ b/libavcodec/arm/Makefile
@@ -59,6 +59,7 @@ ARMV6-OBJS += arm/dsputil_init_armv6.o \
arm/simple_idct_armv6.o \
VFP-OBJS-$(CONFIG_DCA_DECODER) += arm/synth_filter_vfp.o
+VFP-OBJS-$(CONFIG_MDCT) += arm/mdct_vfp.o
VFP-OBJS-$(HAVE_ARMV6) += arm/fmtconvert_vfp.o
NEON-OBJS-$(CONFIG_FFT) += arm/fft_neon.o \
diff --git a/libavcodec/arm/fft_init_arm.c b/libavcodec/arm/fft_init_arm.c
index fe0acc5..a000ea5 100644
--- a/libavcodec/arm/fft_init_arm.c
+++ b/libavcodec/arm/fft_init_arm.c
@@ -26,6 +26,8 @@
void ff_fft_permute_neon(FFTContext *s, FFTComplex *z);
void ff_fft_calc_neon(FFTContext *s, FFTComplex *z);
+void ff_imdct_half_vfp(FFTContext *s, FFTSample *output, const FFTSample *input);
+
void ff_imdct_calc_neon(FFTContext *s, FFTSample *output, const FFTSample *input);
void ff_imdct_half_neon(FFTContext *s, FFTSample *output, const FFTSample *input);
void ff_mdct_calc_neon(FFTContext *s, FFTSample *output, const FFTSample *input);
@@ -48,6 +50,13 @@ av_cold void ff_fft_init_arm(FFTContext *s)
{
int cpu_flags = av_get_cpu_flags();
+ if (have_vfp(cpu_flags)) {
+#if CONFIG_MDCT
+ if (!have_vfpv3(cpu_flags))
+ s->imdct_half = ff_imdct_half_vfp;
+#endif
+ }
+
if (have_neon(cpu_flags)) {
#if CONFIG_FFT
s->fft_permute = ff_fft_permute_neon;
diff --git a/libavcodec/arm/mdct_vfp.S b/libavcodec/arm/mdct_vfp.S
new file mode 100644
index 0000000..0623e96
--- /dev/null
+++ b/libavcodec/arm/mdct_vfp.S
@@ -0,0 +1,205 @@
+/*
+ * Copyright (c) 2013 RISC OS Open Ltd
+ * Author: Ben Avison <bavison@riscosopen.org>
+ *
+ * This file is part of Libav.
+ *
+ * Libav 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.
+ *
+ * Libav 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 Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/arm/asm.S"
+
+CONTEXT .req a1
+ORIGOUT .req a2
+IN .req a3
+OUT .req v1
+REVTAB .req v2
+TCOS .req v3
+TSIN .req v4
+OLDFPSCR .req v5
+J0 .req a2
+J1 .req a4
+J2 .req ip
+J3 .req lr
+
+.macro prerotation_innerloop
+ .set trig_lo, k
+ .set trig_hi, n4 - k - 2
+ .set in_lo, trig_lo * 2
+ .set in_hi, trig_hi * 2
+ vldr d8, [TCOS, #trig_lo*4] @ s16,s17
+ vldr d9, [TCOS, #trig_hi*4] @ s18,s19
+ vldr s0, [IN, #in_hi*4 + 12]
+ vldr s1, [IN, #in_hi*4 + 4]
+ vldr s2, [IN, #in_lo*4 + 12]
+ vldr s3, [IN, #in_lo*4 + 4]
+ vmul.f s8, s0, s16 @ vector operation
+ vldr d10, [TSIN, #trig_lo*4] @ s20,s21
+ vldr d11, [TSIN, #trig_hi*4] @ s22,s23
+ vldr s4, [IN, #in_lo*4]
+ vldr s5, [IN, #in_lo*4 + 8]
+ vldr s6, [IN, #in_hi*4]
+ vldr s7, [IN, #in_hi*4 + 8]
+ ldr J0, [REVTAB, #trig_lo*2]
+ vmul.f s12, s0, s20 @ vector operation
+ ldr J2, [REVTAB, #trig_hi*2]
+ mov J1, J0, lsr #16
+ and J0, J0, #255 @ halfword value will be < n4
+ vmls.f s8, s4, s20 @ vector operation
+ mov J3, J2, lsr #16
+ and J2, J2, #255 @ halfword value will be < n4
+ add J0, OUT, J0, lsl #3
+ vmla.f s12, s4, s16 @ vector operation
+ add J1, OUT, J1, lsl #3
+ add J2, OUT, J2, lsl #3
+ add J3, OUT, J3, lsl #3
+ vstr s8, [J0]
+ vstr s9, [J1]
+ vstr s10, [J2]
+ vstr s11, [J3]
+ vstr s12, [J0, #4]
+ vstr s13, [J1, #4]
+ vstr s14, [J2, #4]
+ vstr s15, [J3, #4]
+ .set k, k + 2
+.endm
+
+.macro postrotation_innerloop tail, head
+ .set trig_lo_head, n8 - k - 2
+ .set trig_hi_head, n8 + k
+ .set out_lo_head, trig_lo_head * 2
+ .set out_hi_head, trig_hi_head * 2
+ .set trig_lo_tail, n8 - (k - 2) - 2
+ .set trig_hi_tail, n8 + (k - 2)
+ .set out_lo_tail, trig_lo_tail * 2
+ .set out_hi_tail, trig_hi_tail * 2
+ .if (k & 2) == 0
+ TCOS_D0_HEAD .req d10 @ s20,s21
+ TCOS_D1_HEAD .req d11 @ s22,s23
+ TCOS_S0_TAIL .req s24
+ .else
+ TCOS_D0_HEAD .req d12 @ s24,s25
+ TCOS_D1_HEAD .req d13 @ s26,s27
+ TCOS_S0_TAIL .req s20
+ .endif
+ .ifnc "\tail",""
+ vmls.f s8, s0, TCOS_S0_TAIL @ vector operation
+ .endif
+ .ifnc "\head",""
+ vldr d8, [TSIN, #trig_lo_head*4] @ s16,s17
+ vldr d9, [TSIN, #trig_hi_head*4] @ s18,s19
+ vldr TCOS_D0_HEAD, [TCOS, #trig_lo_head*4]
+ .endif
+ .ifnc "\tail",""
+ vmla.f s12, s4, TCOS_S0_TAIL @ vector operation
+ .endif
+ .ifnc "\head",""
+ vldr s0, [OUT, #out_lo_head*4]
+ vldr s1, [OUT, #out_lo_head*4 + 8]
+ vldr s2, [OUT, #out_hi_head*4]
+ vldr s3, [OUT, #out_hi_head*4 + 8]
+ vldr s4, [OUT, #out_lo_head*4 + 4]
+ vldr s5, [OUT, #out_lo_head*4 + 12]
+ vldr s6, [OUT, #out_hi_head*4 + 4]
+ vldr s7, [OUT, #out_hi_head*4 + 12]
+ .endif
+ .ifnc "\tail",""
+ vstr s8, [OUT, #out_lo_tail*4]
+ vstr s9, [OUT, #out_lo_tail*4 + 8]
+ vstr s10, [OUT, #out_hi_tail*4]
+ vstr s11, [OUT, #out_hi_tail*4 + 8]
+ .endif
+ .ifnc "\head",""
+ vmul.f s8, s4, s16 @ vector operation
+ .endif
+ .ifnc "\tail",""
+ vstr s12, [OUT, #out_hi_tail*4 + 12]
+ vstr s13, [OUT, #out_hi_tail*4 + 4]
+ vstr s14, [OUT, #out_lo_tail*4 + 12]
+ vstr s15, [OUT, #out_lo_tail*4 + 4]
+ .endif
+ .ifnc "\head",""
+ vmul.f s12, s0, s16 @ vector operation
+ vldr TCOS_D1_HEAD, [TCOS, #trig_hi_head*4]
+ .endif
+ .unreq TCOS_D0_HEAD
+ .unreq TCOS_D1_HEAD
+ .unreq TCOS_S0_TAIL
+ .ifnc "\head",""
+ .set k, k + 2
+ .endif
+.endm
+
+
+/* void ff_imdct_half_vfp(FFTContext *s,
+ * FFTSample *output,
+ * const FFTSample *input)
+ */
+function ff_imdct_half_vfp, export=1
+ ldr ip, [CONTEXT, #5*4] @ mdct_bits
+ teq ip, #6
+ it ne
+ bne ff_imdct_half_c @ only case currently accelerated is the one used by DCA
+
+ .set n, 1<<6
+ .set n2, n/2
+ .set n4, n/4
+ .set n8, n/8
+
+ push {v1-v5,lr}
+ vpush {s16-s27}
+ fmrx OLDFPSCR, FPSCR
+ ldr lr, =0x03030000 @ RunFast mode, short vectors of length 4, stride 1
+ fmxr FPSCR, lr
+ mov OUT, ORIGOUT
+ ldr REVTAB, [CONTEXT, #2*4]
+ ldr TCOS, [CONTEXT, #6*4]
+ ldr TSIN, [CONTEXT, #7*4]
+
+ .set k, 0
+ .rept n8/2
+ prerotation_innerloop
+ .endr
+
+ fmxr FPSCR, OLDFPSCR
+ mov a1, OUT
+ bl ff_fft16_vfp
+ ldr lr, =0x03030000 @ RunFast mode, short vectors of length 4, stride 1
+ fmxr FPSCR, lr
+
+ .set k, 0
+ postrotation_innerloop , head
+ .rept n8/2 - 1
+ postrotation_innerloop tail, head
+ .endr
+ postrotation_innerloop tail
+
+ fmxr FPSCR, OLDFPSCR
+ vpop {s16-s27}
+ pop {v1-v5,pc}
+endfunc
+
+ .unreq CONTEXT
+ .unreq ORIGOUT
+ .unreq IN
+ .unreq OUT
+ .unreq REVTAB
+ .unreq TCOS
+ .unreq TSIN
+ .unreq OLDFPSCR
+ .unreq J0
+ .unreq J1
+ .unreq J2
+ .unreq J3
--
1.7.9.5

58
packages/multimedia/ffmpeg/patches/ffmpeg-0046-ffmpeg-backport-arm-Add-VFP-accelerated-version-of-d.patch
View File

@ -1,58 +0,0 @@
From 8e0babd84c7e03cf678aab8bcf7e2106fe2b3de6 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Martin=20Storsj=C3=B6?= <martin@martin.st>
Date: Fri, 19 Jul 2013 11:03:32 +0300
Subject: [PATCH 45/49] [ffmpeg] - backport - arm: Add VFP-accelerated version
of dca_lfe_fir
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Before After
Mean StdDev Mean StdDev Change
This function 868.2 33.5 436.0 27.0 +99.1%
Overall 15973.0 223.2 15577.5 83.2 +2.5%
Signed-off-by: Martin Storsjö <martin@martin.st>
---
libavcodec/arm/Makefile | 3 ++-
libavcodec/arm/dcadsp_init_arm.c | 4 ++++
2 files changed, 6 insertions(+), 1 deletion(-)
diff --git a/libavcodec/arm/Makefile b/libavcodec/arm/Makefile
index 27e80d5..7fe5bb5 100644
--- a/libavcodec/arm/Makefile
+++ b/libavcodec/arm/Makefile
@@ -58,7 +58,8 @@ ARMV6-OBJS += arm/dsputil_init_armv6.o \
arm/dsputil_armv6.o \
arm/simple_idct_armv6.o \
-VFP-OBJS-$(CONFIG_DCA_DECODER) += arm/synth_filter_vfp.o
+VFP-OBJS-$(CONFIG_DCA_DECODER) += arm/dcadsp_vfp.o \
+ arm/synth_filter_vfp.o
VFP-OBJS-$(CONFIG_MDCT) += arm/mdct_vfp.o
VFP-OBJS-$(HAVE_ARMV6) += arm/fmtconvert_vfp.o
diff --git a/libavcodec/arm/dcadsp_init_arm.c b/libavcodec/arm/dcadsp_init_arm.c
index 56568e0..a1efbff 100644
--- a/libavcodec/arm/dcadsp_init_arm.c
+++ b/libavcodec/arm/dcadsp_init_arm.c
@@ -24,6 +24,8 @@
#include "libavutil/attributes.h"
#include "libavcodec/dcadsp.h"
+void ff_dca_lfe_fir_vfp(float *out, const float *in, const float *coefs,
+ int decifactor, float scale);
void ff_dca_lfe_fir_neon(float *out, const float *in, const float *coefs,
int decifactor, float scale);
@@ -31,6 +33,8 @@ av_cold void ff_dcadsp_init_arm(DCADSPContext *s)
{
int cpu_flags = av_get_cpu_flags();
+ if (have_vfp(cpu_flags) && !have_vfpv3(cpu_flags))
+ s->lfe_fir = ff_dca_lfe_fir_vfp;
if (have_neon(cpu_flags))
s->lfe_fir = ff_dca_lfe_fir_neon;
}
--
1.7.9.5

339
packages/multimedia/ffmpeg/patches/ffmpeg-0047-ffmpeg-backport-arm-Add-VFP-accelerated-version-of-f.patch
View File

@ -1,339 +0,0 @@
From 018b74ea9d8f52788db18ed40838afca05e7b4df Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Martin=20Storsj=C3=B6?= <martin@martin.st>
Date: Fri, 19 Jul 2013 11:23:57 +0300
Subject: [PATCH 46/49] [ffmpeg] - backport - arm: Add VFP-accelerated version
of fft16
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Before After
Mean StdDev Mean StdDev Change
This function 1389.3 4.2 967.8 35.1 +43.6%
Overall 15577.5 83.2 15400.0 336.4 +1.2%
Signed-off-by: Martin Storsjö <martin@martin.st>
---
libavcodec/arm/Makefile | 1 +
libavcodec/arm/fft_vfp.S | 298 +++++++++++++++++++++++++++++++++++
2 files changed, 299 insertions(+)
create mode 100644 libavcodec/arm/fft_vfp.S
diff --git a/libavcodec/arm/Makefile b/libavcodec/arm/Makefile
index 7fe5bb5..7390a8b 100644
--- a/libavcodec/arm/Makefile
+++ b/libavcodec/arm/Makefile
@@ -60,6 +60,7 @@ ARMV6-OBJS += arm/dsputil_init_armv6.o \
VFP-OBJS-$(CONFIG_DCA_DECODER) += arm/dcadsp_vfp.o \
arm/synth_filter_vfp.o
+VFP-OBJS-$(CONFIG_FFT) += arm/fft_vfp.o
VFP-OBJS-$(CONFIG_MDCT) += arm/mdct_vfp.o
VFP-OBJS-$(HAVE_ARMV6) += arm/fmtconvert_vfp.o
diff --git a/libavcodec/arm/fft_vfp.S b/libavcodec/arm/fft_vfp.S
new file mode 100644
index 0000000..7845ebb
--- /dev/null
+++ b/libavcodec/arm/fft_vfp.S
@@ -0,0 +1,298 @@
+/*
+ * Copyright (c) 2013 RISC OS Open Ltd
+ * Author: Ben Avison <bavison@riscosopen.org>
+ *
+ * This file is part of Libav.
+ *
+ * Libav 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.
+ *
+ * Libav 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 Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/arm/asm.S"
+
+@ TODO: * FFTs wider than 16
+@ * dispatch code
+
+function fft4_vfp
+ vldr d0, [a1, #0*2*4] @ s0,s1 = z[0]
+ vldr d4, [a1, #1*2*4] @ s8,s9 = z[1]
+ vldr d1, [a1, #2*2*4] @ s2,s3 = z[2]
+ vldr d5, [a1, #3*2*4] @ s10,s11 = z[3]
+ @ stall
+ vadd.f s12, s0, s8 @ i0
+ vadd.f s13, s1, s9 @ i1
+ vadd.f s14, s2, s10 @ i2
+ vadd.f s15, s3, s11 @ i3
+ vsub.f s8, s0, s8 @ i4
+ vsub.f s9, s1, s9 @ i5
+ vsub.f s10, s2, s10 @ i6
+ vsub.f s11, s3, s11 @ i7
+ @ stall
+ @ stall
+ vadd.f s0, s12, s14 @ z[0].re
+ vsub.f s4, s12, s14 @ z[2].re
+ vadd.f s1, s13, s15 @ z[0].im
+ vsub.f s5, s13, s15 @ z[2].im
+ vadd.f s7, s9, s10 @ z[3].im
+ vsub.f s3, s9, s10 @ z[1].im
+ vadd.f s2, s8, s11 @ z[1].re
+ vsub.f s6, s8, s11 @ z[3].re
+ @ stall
+ @ stall
+ vstr d0, [a1, #0*2*4]
+ vstr d2, [a1, #2*2*4]
+ @ stall
+ @ stall
+ vstr d1, [a1, #1*2*4]
+ vstr d3, [a1, #3*2*4]
+
+ bx lr
+endfunc
+
+.macro macro_fft8_head
+ @ FFT4
+ vldr d4, [a1, #0 * 2*4]
+ vldr d6, [a1, #1 * 2*4]
+ vldr d5, [a1, #2 * 2*4]
+ vldr d7, [a1, #3 * 2*4]
+ @ BF
+ vldr d12, [a1, #4 * 2*4]
+ vadd.f s16, s8, s12 @ vector op
+ vldr d14, [a1, #5 * 2*4]
+ vldr d13, [a1, #6 * 2*4]
+ vldr d15, [a1, #7 * 2*4]
+ vsub.f s20, s8, s12 @ vector op
+ vadd.f s0, s16, s18
+ vsub.f s2, s16, s18
+ vadd.f s1, s17, s19
+ vsub.f s3, s17, s19
+ vadd.f s7, s21, s22
+ vsub.f s5, s21, s22
+ vadd.f s4, s20, s23
+ vsub.f s6, s20, s23
+ vsub.f s20, s24, s28 @ vector op
+ vstr d0, [a1, #0 * 2*4] @ transfer s0-s7 to s24-s31 via memory
+ vstr d1, [a1, #1 * 2*4]
+ vldr s0, cos1pi4
+ vadd.f s16, s24, s28 @ vector op
+ vstr d2, [a1, #2 * 2*4]
+ vstr d3, [a1, #3 * 2*4]
+ vldr d12, [a1, #0 * 2*4]
+ @ TRANSFORM
+ vmul.f s20, s20, s0 @ vector x scalar op
+ vldr d13, [a1, #1 * 2*4]
+ vldr d14, [a1, #2 * 2*4]
+ vldr d15, [a1, #3 * 2*4]
+ @ BUTTERFLIES
+ vadd.f s0, s18, s16
+ vadd.f s1, s17, s19
+ vsub.f s2, s17, s19
+ vsub.f s3, s18, s16
+ vadd.f s4, s21, s20
+ vsub.f s5, s21, s20
+ vadd.f s6, s22, s23
+ vsub.f s7, s22, s23
+ vadd.f s8, s0, s24 @ vector op
+ vstr d0, [a1, #0 * 2*4] @ transfer s0-s3 to s12-s15 via memory
+ vstr d1, [a1, #1 * 2*4]
+ vldr d6, [a1, #0 * 2*4]
+ vldr d7, [a1, #1 * 2*4]
+ vadd.f s1, s5, s6
+ vadd.f s0, s7, s4
+ vsub.f s2, s5, s6
+ vsub.f s3, s7, s4
+ vsub.f s12, s24, s12 @ vector op
+ vsub.f s5, s29, s1
+ vsub.f s4, s28, s0
+ vsub.f s6, s30, s2
+ vsub.f s7, s31, s3
+ vadd.f s16, s0, s28 @ vector op
+ vstr d6, [a1, #4 * 2*4]
+ vstr d7, [a1, #6 * 2*4]
+ vstr d4, [a1, #0 * 2*4]
+ vstr d5, [a1, #2 * 2*4]
+ vstr d2, [a1, #5 * 2*4]
+ vstr d3, [a1, #7 * 2*4]
+.endm
+
+.macro macro_fft8_tail
+ vstr d8, [a1, #1 * 2*4]
+ vstr d9, [a1, #3 * 2*4]
+.endm
+
+function fft8_vfp
+ ldr a3, =0x03030000 @ RunFast mode, vector length 4, stride 1
+ fmrx a2, FPSCR
+ fmxr FPSCR, a3
+ vpush {s16-s31}
+
+ macro_fft8_head
+ macro_fft8_tail
+
+ vpop {s16-s31}
+ fmxr FPSCR, a2
+ bx lr
+endfunc
+
+.align 3
+cos1pi4: @ cos(1*pi/4) = sqrt(2)
+ .float 0.707106769084930419921875
+cos1pi8: @ cos(1*pi/8) = sqrt(2+sqrt(2))/2
+ .float 0.92387950420379638671875
+cos3pi8: @ cos(2*pi/8) = sqrt(2-sqrt(2))/2
+ .float 0.3826834261417388916015625
+
+function ff_fft16_vfp, export=1
+ ldr a3, =0x03030000 @ RunFast mode, vector length 4, stride 1
+ fmrx a2, FPSCR
+ fmxr FPSCR, a3
+ vpush {s16-s31}
+
+ macro_fft8_head
+ @ FFT4(z+8)
+ vldr d10, [a1, #8 * 2*4]
+ vldr d12, [a1, #9 * 2*4]
+ vldr d11, [a1, #10 * 2*4]
+ vldr d13, [a1, #11 * 2*4]
+ macro_fft8_tail
+ vadd.f s16, s20, s24 @ vector op
+ @ FFT4(z+12)
+ vldr d4, [a1, #12 * 2*4]
+ vldr d6, [a1, #13 * 2*4]
+ vldr d5, [a1, #14 * 2*4]
+ vsub.f s20, s20, s24 @ vector op
+ vldr d7, [a1, #15 * 2*4]
+ vadd.f s0, s16, s18
+ vsub.f s4, s16, s18
+ vadd.f s1, s17, s19
+ vsub.f s5, s17, s19
+ vadd.f s7, s21, s22
+ vsub.f s3, s21, s22
+ vadd.f s2, s20, s23
+ vsub.f s6, s20, s23
+ vadd.f s16, s8, s12 @ vector op
+ vstr d0, [a1, #8 * 2*4]
+ vstr d2, [a1, #10 * 2*4]
+ vstr d1, [a1, #9 * 2*4]
+ vsub.f s20, s8, s12
+ vstr d3, [a1, #11 * 2*4]
+ @ TRANSFORM(z[2],z[6],z[10],z[14],cos1pi4,cos1pi4)
+ vldr d12, [a1, #10 * 2*4]
+ vadd.f s0, s16, s18
+ vadd.f s1, s17, s19
+ vsub.f s6, s16, s18
+ vsub.f s7, s17, s19
+ vsub.f s3, s21, s22
+ vadd.f s2, s20, s23
+ vadd.f s5, s21, s22
+ vsub.f s4, s20, s23
+ vstr d0, [a1, #12 * 2*4]
+ vmov s0, s6
+ @ TRANSFORM(z[1],z[5],z[9],z[13],cos1pi8,cos3pi8)
+ vldr d6, [a1, #9 * 2*4]
+ vstr d1, [a1, #13 * 2*4]
+ vldr d1, cos1pi4 @ s2 = cos1pi4, s3 = cos1pi8
+ vstr d2, [a1, #15 * 2*4]
+ vldr d7, [a1, #13 * 2*4]
+ vadd.f s4, s25, s24
+ vsub.f s5, s25, s24
+ vsub.f s6, s0, s7
+ vadd.f s7, s0, s7
+ vmul.f s20, s12, s3 @ vector op
+ @ TRANSFORM(z[3],z[7],z[11],z[15],cos3pi8,cos1pi8)
+ vldr d4, [a1, #11 * 2*4]
+ vldr d5, [a1, #15 * 2*4]
+ vldr s1, cos3pi8
+ vmul.f s24, s4, s2 @ vector * scalar op
+ vmul.f s28, s12, s1 @ vector * scalar op
+ vmul.f s12, s8, s1 @ vector * scalar op
+ vadd.f s4, s20, s29
+ vsub.f s5, s21, s28
+ vsub.f s6, s22, s31
+ vadd.f s7, s23, s30
+ vmul.f s8, s8, s3 @ vector * scalar op
+ vldr d8, [a1, #1 * 2*4]
+ vldr d9, [a1, #5 * 2*4]
+ vldr d10, [a1, #3 * 2*4]
+ vldr d11, [a1, #7 * 2*4]
+ vldr d14, [a1, #2 * 2*4]
+ vadd.f s0, s6, s4
+ vadd.f s1, s5, s7
+ vsub.f s2, s5, s7
+ vsub.f s3, s6, s4
+ vadd.f s4, s12, s9
+ vsub.f s5, s13, s8
+ vsub.f s6, s14, s11
+ vadd.f s7, s15, s10
+ vadd.f s12, s0, s16 @ vector op
+ vstr d0, [a1, #1 * 2*4]
+ vstr d1, [a1, #5 * 2*4]
+ vldr d4, [a1, #1 * 2*4]
+ vldr d5, [a1, #5 * 2*4]
+ vadd.f s0, s6, s4
+ vadd.f s1, s5, s7
+ vsub.f s2, s5, s7
+ vsub.f s3, s6, s4
+ vsub.f s8, s16, s8 @ vector op
+ vstr d6, [a1, #1 * 2*4]
+ vstr d7, [a1, #5 * 2*4]
+ vldr d15, [a1, #6 * 2*4]
+ vsub.f s4, s20, s0
+ vsub.f s5, s21, s1
+ vsub.f s6, s22, s2
+ vsub.f s7, s23, s3
+ vadd.f s20, s0, s20 @ vector op
+ vstr d4, [a1, #9 * 2*4]
+ @ TRANSFORM_ZERO(z[0],z[4],z[8],z[12])
+ vldr d6, [a1, #8 * 2*4]
+ vstr d5, [a1, #13 * 2*4]
+ vldr d7, [a1, #12 * 2*4]
+ vstr d2, [a1, #11 * 2*4]
+ vldr d8, [a1, #0 * 2*4]
+ vstr d3, [a1, #15 * 2*4]
+ vldr d9, [a1, #4 * 2*4]
+ vadd.f s0, s26, s24
+ vadd.f s1, s25, s27
+ vsub.f s2, s25, s27
+ vsub.f s3, s26, s24
+ vadd.f s4, s14, s12
+ vadd.f s5, s13, s15
+ vsub.f s6, s13, s15
+ vsub.f s7, s14, s12
+ vadd.f s8, s0, s28 @ vector op
+ vstr d0, [a1, #3 * 2*4]
+ vstr d1, [a1, #7 * 2*4]
+ vldr d6, [a1, #3 * 2*4]
+ vldr d7, [a1, #7 * 2*4]
+ vsub.f s0, s16, s4
+ vsub.f s1, s17, s5
+ vsub.f s2, s18, s6
+ vsub.f s3, s19, s7
+ vsub.f s12, s28, s12 @ vector op
+ vadd.f s16, s4, s16 @ vector op
+ vstr d10, [a1, #3 * 2*4]
+ vstr d11, [a1, #7 * 2*4]
+ vstr d4, [a1, #2 * 2*4]
+ vstr d5, [a1, #6 * 2*4]
+ vstr d0, [a1, #8 * 2*4]
+ vstr d1, [a1, #12 * 2*4]
+ vstr d6, [a1, #10 * 2*4]
+ vstr d7, [a1, #14 * 2*4]
+ vstr d8, [a1, #0 * 2*4]
+ vstr d9, [a1, #4 * 2*4]
+
+ vpop {s16-s31}
+ fmxr FPSCR, a2
+ bx lr
+endfunc
--
1.7.9.5

140
packages/multimedia/ffmpeg/patches/ffmpeg-0048-ffmpeg-backport-dcadsp-Add-a-new-method-qmf_32_subba.patch
View File

@ -1,140 +0,0 @@
From ed16009b0a05fbd344832d5ad2e982c169aec42c Mon Sep 17 00:00:00 2001
From: Ben Avison <bavison@riscosopen.org>
Date: Mon, 15 Jul 2013 18:28:16 +0100
Subject: [PATCH 47/49] [ffmpeg] - backport - dcadsp: Add a new method,
qmf_32_subbands
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
This does most of the work formerly carried out by
the static function qmf_32_subbands() in dcadec.c.
Signed-off-by: Martin Storsjö <martin@martin.st>
---
libavcodec/dcadec.c | 26 +++++---------------------
libavcodec/dcadsp.c | 30 ++++++++++++++++++++++++++++++
libavcodec/dcadsp.h | 9 +++++++++
3 files changed, 44 insertions(+), 21 deletions(-)
diff --git a/libavcodec/dcadec.c b/libavcodec/dcadec.c
index b648613..4054d63 100644
--- a/libavcodec/dcadec.c
+++ b/libavcodec/dcadec.c
@@ -1108,10 +1108,8 @@ static void qmf_32_subbands(DCAContext *s, int chans,
float scale)
{
const float *prCoeff;
- int i;
int sb_act = s->subband_activity[chans];
- int subindex;
scale *= sqrt(1 / 8.0);
@@ -1121,25 +1119,11 @@ static void qmf_32_subbands(DCAContext *s, int chans,
else /* Perfect reconstruction */
prCoeff = fir_32bands_perfect;
- for (i = sb_act; i < 32; i++)
- s->raXin[i] = 0.0;
-
- /* Reconstructed channel sample index */
- for (subindex = 0; subindex < 8; subindex++) {
- /* Load in one sample from each subband and clear inactive subbands */
- for (i = 0; i < sb_act; i++) {
- unsigned sign = (i - 1) & 2;
- uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ sign << 30;
- AV_WN32A(&s->raXin[i], v);
- }
-
- s->synth.synth_filter_float(&s->imdct,
- s->subband_fir_hist[chans],
- &s->hist_index[chans],
- s->subband_fir_noidea[chans], prCoeff,
- samples_out, s->raXin, scale);
- samples_out += 32;
- }
+ s->dcadsp.qmf_32_subbands(samples_in, sb_act, &s->synth, &s->imdct,
+ s->subband_fir_hist[chans],
+ &s->hist_index[chans],
+ s->subband_fir_noidea[chans], prCoeff,
+ samples_out, s->raXin, scale);
}
static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
diff --git a/libavcodec/dcadsp.c b/libavcodec/dcadsp.c
index dd4994d..ab63f1b 100644
--- a/libavcodec/dcadsp.c
+++ b/libavcodec/dcadsp.c
@@ -20,6 +20,7 @@
*/
#include "config.h"
+#include "libavutil/intreadwrite.h"
#include "dcadsp.h"
static void dca_lfe_fir_c(float *out, const float *in, const float *coefs,
@@ -44,8 +45,37 @@ static void dca_lfe_fir_c(float *out, const float *in, const float *coefs,
}
}
+static void dca_qmf_32_subbands(float samples_in[32][8], int sb_act,
+ SynthFilterContext *synth, FFTContext *imdct,
+ float synth_buf_ptr[512],
+ int *synth_buf_offset, float synth_buf2[32],
+ const float window[512], float *samples_out,
+ float raXin[32], float scale)
+{
+ int i;
+ int subindex;
+
+ for (i = sb_act; i < 32; i++)
+ raXin[i] = 0.0;
+
+ /* Reconstructed channel sample index */
+ for (subindex = 0; subindex < 8; subindex++) {
+ /* Load in one sample from each subband and clear inactive subbands */
+ for (i = 0; i < sb_act; i++) {
+ unsigned sign = (i - 1) & 2;
+ uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ sign << 30;
+ AV_WN32A(&raXin[i], v);
+ }
+
+ synth->synth_filter_float(imdct, synth_buf_ptr, synth_buf_offset,
+ synth_buf2, window, samples_out, raXin, scale);
+ samples_out += 32;
+ }
+}
+
void ff_dcadsp_init(DCADSPContext *s)
{
s->lfe_fir = dca_lfe_fir_c;
+ s->qmf_32_subbands = dca_qmf_32_subbands;
if (ARCH_ARM) ff_dcadsp_init_arm(s);
}
diff --git a/libavcodec/dcadsp.h b/libavcodec/dcadsp.h
index bb157f7..d86c1f3 100644
--- a/libavcodec/dcadsp.h
+++ b/libavcodec/dcadsp.h
@@ -19,9 +19,18 @@
#ifndef AVCODEC_DCADSP_H
#define AVCODEC_DCADSP_H
+#include "avfft.h"
+#include "synth_filter.h"
+
typedef struct DCADSPContext {
void (*lfe_fir)(float *out, const float *in, const float *coefs,
int decifactor, float scale);
+ void (*qmf_32_subbands)(float samples_in[32][8], int sb_act,
+ SynthFilterContext *synth, FFTContext *imdct,
+ float synth_buf_ptr[512],
+ int *synth_buf_offset, float synth_buf2[32],
+ const float window[512], float *samples_out,
+ float raXin[32], float scale);
} DCADSPContext;
void ff_dcadsp_init(DCADSPContext *s);
--
1.7.9.5

551
packages/multimedia/ffmpeg/patches/ffmpeg-0049-ffmpeg-backport-arm-Add-VFP-accelerated-version-of-q.patch
View File

@ -1,551 +0,0 @@
From a6c273927c5bb212e806be6ae10c81dcd81b2152 Mon Sep 17 00:00:00 2001
From: Ben Avison <bavison@riscosopen.org>
Date: Mon, 15 Jul 2013 18:28:17 +0100
Subject: [PATCH 48/49] [ffmpeg] - backport - arm: Add VFP-accelerated version
of qmf_32_subbands
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Before After
Mean StdDev Mean StdDev Change
This function 1323.0 98.0 746.2 60.6 +77.3%
Overall 15400.0 336.4 14147.5 288.4 +8.9%
Signed-off-by: Martin Storsjö <martin@martin.st>
---
libavcodec/arm/dcadsp_init_arm.c | 10 +-
libavcodec/arm/dcadsp_vfp.S | 493 +++++++++++++++++++++++++++
2 files changed, 502 insertions(+), 1 deletion(-)
create mode 100644 libavcodec/arm/dcadsp_vfp.S
diff --git a/libavcodec/arm/dcadsp_init_arm.c b/libavcodec/arm/dcadsp_init_arm.c
index a1efbff..58267a2 100644
--- a/libavcodec/arm/dcadsp_init_arm.c
+++ b/libavcodec/arm/dcadsp_init_arm.c
@@ -26,6 +26,12 @@
void ff_dca_lfe_fir_vfp(float *out, const float *in, const float *coefs,
int decifactor, float scale);
+void ff_dca_qmf_32_subbands_vfp(float samples_in[32][8], int sb_act,
+ SynthFilterContext *synth, FFTContext *imdct,
+ float synth_buf_ptr[512],
+ int *synth_buf_offset, float synth_buf2[32],
+ const float window[512], float *samples_out,
+ float raXin[32], float scale);
void ff_dca_lfe_fir_neon(float *out, const float *in, const float *coefs,
int decifactor, float scale);
@@ -33,8 +39,10 @@ av_cold void ff_dcadsp_init_arm(DCADSPContext *s)
{
int cpu_flags = av_get_cpu_flags();
- if (have_vfp(cpu_flags) && !have_vfpv3(cpu_flags))
+ if (have_vfp(cpu_flags) && !have_vfpv3(cpu_flags)) {
s->lfe_fir = ff_dca_lfe_fir_vfp;
+ s->qmf_32_subbands = ff_dca_qmf_32_subbands_vfp;
+ }
if (have_neon(cpu_flags))
s->lfe_fir = ff_dca_lfe_fir_neon;
}
diff --git a/libavcodec/arm/dcadsp_vfp.S b/libavcodec/arm/dcadsp_vfp.S
new file mode 100644
index 0000000..6039e87
--- /dev/null
+++ b/libavcodec/arm/dcadsp_vfp.S
@@ -0,0 +1,493 @@
+/*
+ * Copyright (c) 2013 RISC OS Open Ltd
+ * Author: Ben Avison <bavison@riscosopen.org>
+ *
+ * This file is part of Libav.
+ *
+ * Libav 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.
+ *
+ * Libav 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 Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/arm/asm.S"
+
+POUT .req a1
+PIN .req a2
+PCOEF .req a3
+DECIFACTOR .req a4
+OLDFPSCR .req a4
+COUNTER .req ip
+
+SCALE32 .req s28 @ use vector of 4 in place of 9th scalar when decifactor=32 / JMAX=8
+SCALE64 .req s0 @ spare register in scalar bank when decifactor=64 / JMAX=4
+IN0 .req s4
+IN1 .req s5
+IN2 .req s6
+IN3 .req s7
+IN4 .req s0
+IN5 .req s1
+IN6 .req s2
+IN7 .req s3
+COEF0 .req s8 @ coefficient elements
+COEF1 .req s9
+COEF2 .req s10
+COEF3 .req s11
+COEF4 .req s12
+COEF5 .req s13
+COEF6 .req s14
+COEF7 .req s15
+ACCUM0 .req s16 @ double-buffered multiply-accumulate results
+ACCUM4 .req s20
+POST0 .req s24 @ do long-latency post-multiply in this vector in parallel
+POST1 .req s25
+POST2 .req s26
+POST3 .req s27
+
+
+.macro inner_loop decifactor, dir, tail, head
+ .ifc "\dir","up"
+ .set X, 0
+ .set Y, 4
+ .else
+ .set X, 4*JMAX*4 - 4
+ .set Y, -4
+ .endif
+ .ifnc "\head",""
+ vldr COEF0, [PCOEF, #X + (0*JMAX + 0) * Y]
+ vldr COEF1, [PCOEF, #X + (1*JMAX + 0) * Y]
+ vldr COEF2, [PCOEF, #X + (2*JMAX + 0) * Y]
+ vldr COEF3, [PCOEF, #X + (3*JMAX + 0) * Y]
+ .endif
+ .ifnc "\tail",""
+ vadd.f POST0, ACCUM0, ACCUM4 @ vector operation
+ .endif
+ .ifnc "\head",""
+ vmul.f ACCUM0, COEF0, IN0 @ vector = vector * scalar
+ vldr COEF4, [PCOEF, #X + (0*JMAX + 1) * Y]
+ vldr COEF5, [PCOEF, #X + (1*JMAX + 1) * Y]
+ vldr COEF6, [PCOEF, #X + (2*JMAX + 1) * Y]
+ .endif
+ .ifnc "\tail",""
+ vmul.f POST0, POST0, SCALE\decifactor @ vector operation (SCALE may be scalar)
+ .endif
+ .ifnc "\head",""
+ vldr COEF7, [PCOEF, #X + (3*JMAX + 1) * Y]
+ .ifc "\tail",""
+ vmul.f ACCUM4, COEF4, IN1 @ vector operation
+ .endif
+ vldr COEF0, [PCOEF, #X + (0*JMAX + 2) * Y]
+ vldr COEF1, [PCOEF, #X + (1*JMAX + 2) * Y]
+ .ifnc "\tail",""
+ vmul.f ACCUM4, COEF4, IN1 @ vector operation
+ .endif
+ vldr COEF2, [PCOEF, #X + (2*JMAX + 2) * Y]
+ vldr COEF3, [PCOEF, #X + (3*JMAX + 2) * Y]
+ .endif
+ .ifnc "\tail",""
+ vstmia POUT!, {POST0-POST3}
+ .endif
+ .ifnc "\head",""
+ vmla.f ACCUM0, COEF0, IN2 @ vector = vector * scalar
+ vldr COEF4, [PCOEF, #X + (0*JMAX + 3) * Y]
+ vldr COEF5, [PCOEF, #X + (1*JMAX + 3) * Y]
+ vldr COEF6, [PCOEF, #X + (2*JMAX + 3) * Y]
+ vldr COEF7, [PCOEF, #X + (3*JMAX + 3) * Y]
+ vmla.f ACCUM4, COEF4, IN3 @ vector = vector * scalar
+ .if \decifactor == 32
+ vldr COEF0, [PCOEF, #X + (0*JMAX + 4) * Y]
+ vldr COEF1, [PCOEF, #X + (1*JMAX + 4) * Y]
+ vldr COEF2, [PCOEF, #X + (2*JMAX + 4) * Y]
+ vldr COEF3, [PCOEF, #X + (3*JMAX + 4) * Y]
+ vmla.f ACCUM0, COEF0, IN4 @ vector = vector * scalar
+ vldr COEF4, [PCOEF, #X + (0*JMAX + 5) * Y]
+ vldr COEF5, [PCOEF, #X + (1*JMAX + 5) * Y]
+ vldr COEF6, [PCOEF, #X + (2*JMAX + 5) * Y]
+ vldr COEF7, [PCOEF, #X + (3*JMAX + 5) * Y]
+ vmla.f ACCUM4, COEF4, IN5 @ vector = vector * scalar
+ vldr COEF0, [PCOEF, #X + (0*JMAX + 6) * Y]
+ vldr COEF1, [PCOEF, #X + (1*JMAX + 6) * Y]
+ vldr COEF2, [PCOEF, #X + (2*JMAX + 6) * Y]
+ vldr COEF3, [PCOEF, #X + (3*JMAX + 6) * Y]
+ vmla.f ACCUM0, COEF0, IN6 @ vector = vector * scalar
+ vldr COEF4, [PCOEF, #X + (0*JMAX + 7) * Y]
+ vldr COEF5, [PCOEF, #X + (1*JMAX + 7) * Y]
+ vldr COEF6, [PCOEF, #X + (2*JMAX + 7) * Y]
+ vldr COEF7, [PCOEF, #X + (3*JMAX + 7) * Y]
+ vmla.f ACCUM4, COEF4, IN7 @ vector = vector * scalar
+ .endif
+ .endif
+.endm
+
+.macro dca_lfe_fir decifactor
+ .if \decifactor == 32
+ .set JMAX, 8
+ vpush {s16-s31}
+ vmov SCALE32, s0 @ duplicate scalar across vector
+ vldr IN4, [PIN, #-4*4]
+ vldr IN5, [PIN, #-5*4]
+ vldr IN6, [PIN, #-6*4]
+ vldr IN7, [PIN, #-7*4]
+ .else
+ .set JMAX, 4
+ vpush {s16-s27}
+ .endif
+
+ mov COUNTER, #\decifactor/4 - 1
+ inner_loop \decifactor, up,, head
+1: add PCOEF, PCOEF, #4*JMAX*4
+ subs COUNTER, COUNTER, #1
+ inner_loop \decifactor, up, tail, head
+ bne 1b
+ inner_loop \decifactor, up, tail
+
+ mov COUNTER, #\decifactor/4 - 1
+ inner_loop \decifactor, down,, head
+1: sub PCOEF, PCOEF, #4*JMAX*4
+ subs COUNTER, COUNTER, #1
+ inner_loop \decifactor, down, tail, head
+ bne 1b
+ inner_loop \decifactor, down, tail
+
+ .if \decifactor == 32
+ vpop {s16-s31}
+ .else
+ vpop {s16-s27}
+ .endif
+ fmxr FPSCR, OLDFPSCR
+ bx lr
+.endm
+
+
+/* void ff_dca_lfe_fir_vfp(float *out, const float *in, const float *coefs,
+ * int decifactor, float scale)
+ */
+function ff_dca_lfe_fir_vfp, export=1
+ teq DECIFACTOR, #32
+ fmrx OLDFPSCR, FPSCR
+ ldr ip, =0x03030000 @ RunFast mode, short vectors of length 4, stride 1
+ fmxr FPSCR, ip
+NOVFP vldr s0, [sp]
+ vldr IN0, [PIN, #-0*4]
+ vldr IN1, [PIN, #-1*4]
+ vldr IN2, [PIN, #-2*4]
+ vldr IN3, [PIN, #-3*4]
+ beq 32f
+64: dca_lfe_fir 64
+ .ltorg
+32: dca_lfe_fir 32
+endfunc
+
+ .unreq POUT
+ .unreq PIN
+ .unreq PCOEF
+ .unreq DECIFACTOR
+ .unreq OLDFPSCR
+ .unreq COUNTER
+
+ .unreq SCALE32
+ .unreq SCALE64
+ .unreq IN0
+ .unreq IN1
+ .unreq IN2
+ .unreq IN3
+ .unreq IN4
+ .unreq IN5
+ .unreq IN6
+ .unreq IN7
+ .unreq COEF0
+ .unreq COEF1
+ .unreq COEF2
+ .unreq COEF3
+ .unreq COEF4
+ .unreq COEF5
+ .unreq COEF6
+ .unreq COEF7
+ .unreq ACCUM0
+ .unreq ACCUM4
+ .unreq POST0
+ .unreq POST1
+ .unreq POST2
+ .unreq POST3
+
+
+IN .req a1
+SBACT .req a2
+OLDFPSCR .req a3
+IMDCT .req a4
+WINDOW .req v1
+OUT .req v2
+BUF .req v3
+SCALEINT .req v4 @ only used in softfp case
+COUNT .req v5
+
+SCALE .req s0
+
+/* Stack layout differs in softfp and hardfp cases:
+ *
+ * hardfp
+ * fp -> 6 arg words saved by caller
+ * a3,a4,v1-v3,v5,fp,lr on entry (a3 just to pad to 8 bytes)
+ * s16-s23 on entry
+ * align 16
+ * buf -> 8*32*4 bytes buffer
+ * s0 on entry
+ * sp -> 3 arg words for callee
+ *
+ * softfp
+ * fp -> 7 arg words saved by caller
+ * a4,v1-v5,fp,lr on entry
+ * s16-s23 on entry
+ * align 16
+ * buf -> 8*32*4 bytes buffer
+ * sp -> 4 arg words for callee
+ */
+
+/* void ff_dca_qmf_32_subbands_vfp(float samples_in[32][8], int sb_act,
+ * SynthFilterContext *synth, FFTContext *imdct,
+ * float (*synth_buf_ptr)[512],
+ * int *synth_buf_offset, float (*synth_buf2)[32],
+ * const float (*window)[512], float *samples_out,
+ * float (*raXin)[32], float scale);
+ */
+function ff_dca_qmf_32_subbands_vfp, export=1
+VFP push {a3-a4,v1-v3,v5,fp,lr}
+NOVFP push {a4,v1-v5,fp,lr}
+ add fp, sp, #8*4
+ vpush {s16-s23}
+ @ The buffer pointed at by raXin isn't big enough for us to do a
+ @ complete matrix transposition as we want to, so allocate an
+ @ alternative buffer from the stack. Align to 4 words for speed.
+ sub BUF, sp, #8*32*4
+ bic BUF, BUF, #15
+ mov sp, BUF
+ ldr lr, =0x03330000 @ RunFast mode, short vectors of length 4, stride 2
+ fmrx OLDFPSCR, FPSCR
+ fmxr FPSCR, lr
+ @ COUNT is used to count down 2 things at once:
+ @ bits 0-4 are the number of word pairs remaining in the output row
+ @ bits 5-31 are the number of words to copy (with possible negation)
+ @ from the source matrix before we start zeroing the remainder
+ mov COUNT, #(-4 << 5) + 16
+ adds COUNT, COUNT, SBACT, lsl #5
+ bmi 2f
+1:
+ vldr s8, [IN, #(0*8+0)*4]
+ vldr s10, [IN, #(0*8+1)*4]
+ vldr s12, [IN, #(0*8+2)*4]
+ vldr s14, [IN, #(0*8+3)*4]
+ vldr s16, [IN, #(0*8+4)*4]
+ vldr s18, [IN, #(0*8+5)*4]
+ vldr s20, [IN, #(0*8+6)*4]
+ vldr s22, [IN, #(0*8+7)*4]
+ vneg.f s8, s8
+ vldr s9, [IN, #(1*8+0)*4]
+ vldr s11, [IN, #(1*8+1)*4]
+ vldr s13, [IN, #(1*8+2)*4]
+ vldr s15, [IN, #(1*8+3)*4]
+ vneg.f s16, s16
+ vldr s17, [IN, #(1*8+4)*4]
+ vldr s19, [IN, #(1*8+5)*4]
+ vldr s21, [IN, #(1*8+6)*4]
+ vldr s23, [IN, #(1*8+7)*4]
+ vstr d4, [BUF, #(0*32+0)*4]
+ vstr d5, [BUF, #(1*32+0)*4]
+ vstr d6, [BUF, #(2*32+0)*4]
+ vstr d7, [BUF, #(3*32+0)*4]
+ vstr d8, [BUF, #(4*32+0)*4]
+ vstr d9, [BUF, #(5*32+0)*4]
+ vstr d10, [BUF, #(6*32+0)*4]
+ vstr d11, [BUF, #(7*32+0)*4]
+ vldr s9, [IN, #(3*8+0)*4]
+ vldr s11, [IN, #(3*8+1)*4]
+ vldr s13, [IN, #(3*8+2)*4]
+ vldr s15, [IN, #(3*8+3)*4]
+ vldr s17, [IN, #(3*8+4)*4]
+ vldr s19, [IN, #(3*8+5)*4]
+ vldr s21, [IN, #(3*8+6)*4]
+ vldr s23, [IN, #(3*8+7)*4]
+ vneg.f s9, s9
+ vldr s8, [IN, #(2*8+0)*4]
+ vldr s10, [IN, #(2*8+1)*4]
+ vldr s12, [IN, #(2*8+2)*4]
+ vldr s14, [IN, #(2*8+3)*4]
+ vneg.f s17, s17
+ vldr s16, [IN, #(2*8+4)*4]
+ vldr s18, [IN, #(2*8+5)*4]
+ vldr s20, [IN, #(2*8+6)*4]
+ vldr s22, [IN, #(2*8+7)*4]
+ vstr d4, [BUF, #(0*32+2)*4]
+ vstr d5, [BUF, #(1*32+2)*4]
+ vstr d6, [BUF, #(2*32+2)*4]
+ vstr d7, [BUF, #(3*32+2)*4]
+ vstr d8, [BUF, #(4*32+2)*4]
+ vstr d9, [BUF, #(5*32+2)*4]
+ vstr d10, [BUF, #(6*32+2)*4]
+ vstr d11, [BUF, #(7*32+2)*4]
+ add IN, IN, #4*8*4
+ add BUF, BUF, #4*4
+ subs COUNT, COUNT, #(4 << 5) + 2
+ bpl 1b
+2: @ Now deal with trailing < 4 samples
+ adds COUNT, COUNT, #3 << 5
+ bmi 4f @ sb_act was a multiple of 4
+ bics lr, COUNT, #0x1F
+ bne 3f
+ @ sb_act was n*4+1
+ vldr s8, [IN, #(0*8+0)*4]
+ vldr s10, [IN, #(0*8+1)*4]
+ vldr s12, [IN, #(0*8+2)*4]
+ vldr s14, [IN, #(0*8+3)*4]
+ vldr s16, [IN, #(0*8+4)*4]
+ vldr s18, [IN, #(0*8+5)*4]
+ vldr s20, [IN, #(0*8+6)*4]
+ vldr s22, [IN, #(0*8+7)*4]
+ vneg.f s8, s8
+ vldr s9, zero
+ vldr s11, zero
+ vldr s13, zero
+ vldr s15, zero
+ vneg.f s16, s16
+ vldr s17, zero
+ vldr s19, zero
+ vldr s21, zero
+ vldr s23, zero
+ vstr d4, [BUF, #(0*32+0)*4]
+ vstr d5, [BUF, #(1*32+0)*4]
+ vstr d6, [BUF, #(2*32+0)*4]
+ vstr d7, [BUF, #(3*32+0)*4]
+ vstr d8, [BUF, #(4*32+0)*4]
+ vstr d9, [BUF, #(5*32+0)*4]
+ vstr d10, [BUF, #(6*32+0)*4]
+ vstr d11, [BUF, #(7*32+0)*4]
+ add BUF, BUF, #2*4
+ sub COUNT, COUNT, #1
+ b 4f
+3: @ sb_act was n*4+2 or n*4+3, so do the first 2
+ vldr s8, [IN, #(0*8+0)*4]
+ vldr s10, [IN, #(0*8+1)*4]
+ vldr s12, [IN, #(0*8+2)*4]
+ vldr s14, [IN, #(0*8+3)*4]
+ vldr s16, [IN, #(0*8+4)*4]
+ vldr s18, [IN, #(0*8+5)*4]
+ vldr s20, [IN, #(0*8+6)*4]
+ vldr s22, [IN, #(0*8+7)*4]
+ vneg.f s8, s8
+ vldr s9, [IN, #(1*8+0)*4]
+ vldr s11, [IN, #(1*8+1)*4]
+ vldr s13, [IN, #(1*8+2)*4]
+ vldr s15, [IN, #(1*8+3)*4]
+ vneg.f s16, s16
+ vldr s17, [IN, #(1*8+4)*4]
+ vldr s19, [IN, #(1*8+5)*4]
+ vldr s21, [IN, #(1*8+6)*4]
+ vldr s23, [IN, #(1*8+7)*4]
+ vstr d4, [BUF, #(0*32+0)*4]
+ vstr d5, [BUF, #(1*32+0)*4]
+ vstr d6, [BUF, #(2*32+0)*4]
+ vstr d7, [BUF, #(3*32+0)*4]
+ vstr d8, [BUF, #(4*32+0)*4]
+ vstr d9, [BUF, #(5*32+0)*4]
+ vstr d10, [BUF, #(6*32+0)*4]
+ vstr d11, [BUF, #(7*32+0)*4]
+ add BUF, BUF, #2*4
+ sub COUNT, COUNT, #(2 << 5) + 1
+ bics lr, COUNT, #0x1F
+ bne 4f
+ @ sb_act was n*4+3
+ vldr s8, [IN, #(2*8+0)*4]
+ vldr s10, [IN, #(2*8+1)*4]
+ vldr s12, [IN, #(2*8+2)*4]
+ vldr s14, [IN, #(2*8+3)*4]
+ vldr s16, [IN, #(2*8+4)*4]
+ vldr s18, [IN, #(2*8+5)*4]
+ vldr s20, [IN, #(2*8+6)*4]
+ vldr s22, [IN, #(2*8+7)*4]
+ vldr s9, zero
+ vldr s11, zero
+ vldr s13, zero
+ vldr s15, zero
+ vldr s17, zero
+ vldr s19, zero
+ vldr s21, zero
+ vldr s23, zero
+ vstr d4, [BUF, #(0*32+0)*4]
+ vstr d5, [BUF, #(1*32+0)*4]
+ vstr d6, [BUF, #(2*32+0)*4]
+ vstr d7, [BUF, #(3*32+0)*4]
+ vstr d8, [BUF, #(4*32+0)*4]
+ vstr d9, [BUF, #(5*32+0)*4]
+ vstr d10, [BUF, #(6*32+0)*4]
+ vstr d11, [BUF, #(7*32+0)*4]
+ add BUF, BUF, #2*4
+ sub COUNT, COUNT, #1
+4: @ Now fill the remainder with 0
+ vldr s8, zero
+ vldr s9, zero
+ ands COUNT, COUNT, #0x1F
+ beq 6f
+5: vstr d4, [BUF, #(0*32+0)*4]
+ vstr d4, [BUF, #(1*32+0)*4]
+ vstr d4, [BUF, #(2*32+0)*4]
+ vstr d4, [BUF, #(3*32+0)*4]
+ vstr d4, [BUF, #(4*32+0)*4]
+ vstr d4, [BUF, #(5*32+0)*4]
+ vstr d4, [BUF, #(6*32+0)*4]
+ vstr d4, [BUF, #(7*32+0)*4]
+ add BUF, BUF, #2*4
+ subs COUNT, COUNT, #1
+ bne 5b
+6:
+ fmxr FPSCR, OLDFPSCR
+ ldr WINDOW, [fp, #3*4]
+ ldr OUT, [fp, #4*4]
+ sub BUF, BUF, #32*4
+NOVFP ldr SCALEINT, [fp, #6*4]
+ mov COUNT, #8
+VFP vpush {SCALE}
+VFP sub sp, sp, #3*4
+NOVFP sub sp, sp, #4*4
+7:
+VFP ldr a1, [fp, #-7*4] @ imdct
+NOVFP ldr a1, [fp, #-8*4]
+ ldmia fp, {a2-a4}
+VFP stmia sp, {WINDOW, OUT, BUF}
+NOVFP stmia sp, {WINDOW, OUT, BUF, SCALEINT}
+VFP vldr SCALE, [sp, #3*4]
+ bl ff_synth_filter_float_vfp
+ add OUT, OUT, #32*4
+ add BUF, BUF, #32*4
+ subs COUNT, COUNT, #1
+ bne 7b
+
+A sub sp, fp, #(8+8)*4
+T sub fp, fp, #(8+8)*4
+T mov sp, fp
+ vpop {s16-s23}
+VFP pop {a3-a4,v1-v3,v5,fp,pc}
+NOVFP pop {a4,v1-v5,fp,pc}
+endfunc
+
+ .unreq IN
+ .unreq SBACT
+ .unreq OLDFPSCR
+ .unreq IMDCT
+ .unreq WINDOW
+ .unreq OUT
+ .unreq BUF
+ .unreq SCALEINT
+ .unreq COUNT
+
+ .unreq SCALE
+
+ .align 2
+zero: .word 0
--
1.7.9.5

64
packages/multimedia/ffmpeg/patches/ffmpeg-0050-ffmpeg-backport-arm-Mangle-external-symbols-properly.patch
View File

@ -1,64 +0,0 @@
From 101f5a2c5db12605c24fe4aa41b3fabacfd3bad3 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Martin=20Storsj=C3=B6?= <martin@martin.st>
Date: Mon, 22 Jul 2013 12:33:22 +0300
Subject: [PATCH 49/49] [ffmpeg] - backport - arm: Mangle external symbols
properly in new vfp assembly files
Reviewed-by: Kostya Shishkov
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
---
libavcodec/arm/dcadsp_vfp.S | 2 +-
libavcodec/arm/mdct_vfp.S | 4 ++--
libavcodec/arm/synth_filter_vfp.S | 2 +-
3 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/libavcodec/arm/dcadsp_vfp.S b/libavcodec/arm/dcadsp_vfp.S
index 6039e87..5892a84 100644
--- a/libavcodec/arm/dcadsp_vfp.S
+++ b/libavcodec/arm/dcadsp_vfp.S
@@ -463,7 +463,7 @@ NOVFP ldr a1, [fp, #-8*4]
VFP stmia sp, {WINDOW, OUT, BUF}
NOVFP stmia sp, {WINDOW, OUT, BUF, SCALEINT}
VFP vldr SCALE, [sp, #3*4]
- bl ff_synth_filter_float_vfp
+ bl X(ff_synth_filter_float_vfp)
add OUT, OUT, #32*4
add BUF, BUF, #32*4
subs COUNT, COUNT, #1
diff --git a/libavcodec/arm/mdct_vfp.S b/libavcodec/arm/mdct_vfp.S
index 0623e96..94db24f 100644
--- a/libavcodec/arm/mdct_vfp.S
+++ b/libavcodec/arm/mdct_vfp.S
@@ -151,7 +151,7 @@ function ff_imdct_half_vfp, export=1
ldr ip, [CONTEXT, #5*4] @ mdct_bits
teq ip, #6
it ne
- bne ff_imdct_half_c @ only case currently accelerated is the one used by DCA
+ bne X(ff_imdct_half_c) @ only case currently accelerated is the one used by DCA
.set n, 1<<6
.set n2, n/2
@@ -175,7 +175,7 @@ function ff_imdct_half_vfp, export=1
fmxr FPSCR, OLDFPSCR
mov a1, OUT
- bl ff_fft16_vfp
+ bl X(ff_fft16_vfp)
ldr lr, =0x03030000 @ RunFast mode, short vectors of length 4, stride 1
fmxr FPSCR, lr
diff --git a/libavcodec/arm/synth_filter_vfp.S b/libavcodec/arm/synth_filter_vfp.S
index c219c41..e6e6408 100644
--- a/libavcodec/arm/synth_filter_vfp.S
+++ b/libavcodec/arm/synth_filter_vfp.S
@@ -132,7 +132,7 @@ function ff_synth_filter_float_vfp, export=1
str lr, [P_SB_OFF] @ rotate offset, modulo buffer size, ready for next call
ldr a3, [sp, #(16+6+2)*4] @ fetch in from stack, to pass to imdct_half
VFP vmov s16, SCALE @ imdct_half is free to corrupt s0, but it contains one of our arguments in hardfp case
- bl ff_imdct_half_vfp
+ bl X(ff_imdct_half_vfp)
VFP vmov SCALE, s16
fmrx OLDFPSCR, FPSCR
--
1.7.9.5

72
packages/multimedia/ffmpeg/patches/ffmpeg-0051-ffmpeg-backport-avio-Add-an-internal-function-for-re.patch
View File

@ -1,72 +0,0 @@
From 5ce8f2bf354b7adf904ac3e1438915586c5a0bb1 Mon Sep 17 00:00:00 2001
From: Ben Avison <bavison@riscosopen.org>
Date: Wed, 31 Jul 2013 23:46:08 +0100
Subject: [PATCH 51/54] [ffmpeg] - backport - avio: Add an internal function
for reading without copying
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
As long as there is enough contiguous data in the avio buffer,
just return a pointer to it instead of copying it to the caller
provided buffer.
Signed-off-by: Martin Storsjö <martin@martin.st>
---
libavformat/avio_internal.h | 17 +++++++++++++++++
libavformat/aviobuf.c | 12 ++++++++++++
2 files changed, 29 insertions(+)
diff --git a/libavformat/avio_internal.h b/libavformat/avio_internal.h
index cf36764..e9ece57 100644
--- a/libavformat/avio_internal.h
+++ b/libavformat/avio_internal.h
@@ -38,6 +38,23 @@ int ffio_init_context(AVIOContext *s,
/**
+ * Read size bytes from AVIOContext, returning a pointer.
+ * Note that the data pointed at by the returned pointer is only
+ * valid until the next call that references the same IO context.
+ * @param s IO context
+ * @param buf pointer to buffer into which to assemble the requested
+ * data if it is not available in contiguous addresses in the
+ * underlying buffer
+ * @param size number of bytes requested
+ * @param data address at which to store pointer: this will be a
+ * a direct pointer into the underlying buffer if the requested
+ * number of bytes are available at contiguous addresses, otherwise
+ * will be a copy of buf
+ * @return number of bytes read or AVERROR
+ */
+int ffio_read_indirect(AVIOContext *s, unsigned char *buf, int size, unsigned char **data);
+
+/**
* Read size bytes from AVIOContext into buf.
* This reads at most 1 packet. If that is not enough fewer bytes will be
* returned.
diff --git a/libavformat/aviobuf.c b/libavformat/aviobuf.c
index 7a73a17..465c46d 100644
--- a/libavformat/aviobuf.c
+++ b/libavformat/aviobuf.c
@@ -522,6 +522,18 @@ int avio_read(AVIOContext *s, unsigned char *buf, int size)
return size1 - size;
}
+int ffio_read_indirect(AVIOContext *s, unsigned char *buf, int size, unsigned char **data)
+{
+ if (s->buf_end - s->buf_ptr >= size && !s->write_flag) {
+ *data = s->buf_ptr;
+ s->buf_ptr += size;
+ return size;
+ } else {
+ *data = buf;
+ return avio_read(s, buf, size);
+ }
+}
+
int ffio_read_partial(AVIOContext *s, unsigned char *buf, int size)
{
int len;
--
1.7.9.5

149
packages/multimedia/ffmpeg/patches/ffmpeg-0052-ffmpeg-backport-mpegts-Remove-one-memcpy-per-packet.patch
View File

@ -1,149 +0,0 @@
From 1496d8c12075c0f3783e348a5d73fef9e3000b0f Mon Sep 17 00:00:00 2001
From: Ben Avison <bavison@riscosopen.org>
Date: Wed, 31 Jul 2013 23:46:08 +0100
Subject: [PATCH 52/54] [ffmpeg] - backport - mpegts: Remove one memcpy per
packet
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
This was being performed to ensure that a complete packet was held in
contiguous memory, prior to parsing the packet. However, the source buffer
is typically large enough that the packet was already contiguous, so it is
beneficial to return the packet by reference in most cases.
Before After
Mean StdDev Mean StdDev Change
memcpy 720.7 32.7 649.8 25.1 +10.9%
Overall 2372.7 46.1 2291.7 21.8 +3.5%
Signed-off-by: Martin Storsjö <martin@martin.st>
---
libavformat/mpegts.c | 41 ++++++++++++++++++++++++++-------------
1 file changed, 28 insertions(+), 13 deletions(-)
diff --git a/libavformat/mpegts.c b/libavformat/mpegts.c
index b5f5d63..5307521 100644
--- a/libavformat/mpegts.c
+++ b/libavformat/mpegts.c
@@ -1863,17 +1863,17 @@ static int mpegts_resync(AVFormatContext *s)
}
/* return -1 if error or EOF. Return 0 if OK. */
-static int read_packet(AVFormatContext *s, uint8_t *buf, int raw_packet_size)
+static int read_packet(AVFormatContext *s, uint8_t *buf, int raw_packet_size, uint8_t **data)
{
AVIOContext *pb = s->pb;
- int skip, len;
+ int len;
for(;;) {
- len = avio_read(pb, buf, TS_PACKET_SIZE);
+ len = ffio_read_indirect(pb, buf, TS_PACKET_SIZE, data);
if (len != TS_PACKET_SIZE)
return len < 0 ? len : AVERROR_EOF;
/* check packet sync byte */
- if (buf[0] != 0x47) {
+ if ((*data)[0] != 0x47) {
/* find a new packet start */
avio_seek(pb, -TS_PACKET_SIZE, SEEK_CUR);
if (mpegts_resync(s) < 0)
@@ -1881,19 +1881,25 @@ static int read_packet(AVFormatContext *s, uint8_t *buf, int raw_packet_size)
else
continue;
} else {
- skip = raw_packet_size - TS_PACKET_SIZE;
- if (skip > 0)
- avio_skip(pb, skip);
break;
}
}
return 0;
}
+static void finished_reading_packet(AVFormatContext *s, int raw_packet_size)
+{
+ AVIOContext *pb = s->pb;
+ int skip = raw_packet_size - TS_PACKET_SIZE;
+ if (skip > 0)
+ avio_skip(pb, skip);
+}
+
static int handle_packets(MpegTSContext *ts, int nb_packets)
{
AVFormatContext *s = ts->stream;
uint8_t packet[TS_PACKET_SIZE + FF_INPUT_BUFFER_PADDING_SIZE];
+ uint8_t *data;
int packet_num, ret = 0;
if (avio_tell(s->pb) != ts->last_pos) {
@@ -1926,10 +1932,11 @@ static int handle_packets(MpegTSContext *ts, int nb_packets)
if (ts->stop_parse > 0)
break;
- ret = read_packet(s, packet, ts->raw_packet_size);
+ ret = read_packet(s, packet, ts->raw_packet_size, &data);
if (ret != 0)
break;
- ret = handle_packet(ts, packet);
+ ret = handle_packet(ts, data);
+ finished_reading_packet(s, ts->raw_packet_size);
if (ret != 0)
break;
}
@@ -2087,6 +2094,7 @@ static int mpegts_read_header(AVFormatContext *s)
int64_t pcrs[2], pcr_h;
int packet_count[2];
uint8_t packet[TS_PACKET_SIZE];
+ uint8_t *data;
/* only read packets */
@@ -2102,18 +2110,21 @@ static int mpegts_read_header(AVFormatContext *s)
nb_pcrs = 0;
nb_packets = 0;
for(;;) {
- ret = read_packet(s, packet, ts->raw_packet_size);
+ ret = read_packet(s, packet, ts->raw_packet_size, &data);
if (ret < 0)
return -1;
- pid = AV_RB16(packet + 1) & 0x1fff;
+ pid = AV_RB16(data + 1) & 0x1fff;
if ((pcr_pid == -1 || pcr_pid == pid) &&
- parse_pcr(&pcr_h, &pcr_l, packet) == 0) {
+ parse_pcr(&pcr_h, &pcr_l, data) == 0) {
+ finished_reading_packet(s, ts->raw_packet_size);
pcr_pid = pid;
packet_count[nb_pcrs] = nb_packets;
pcrs[nb_pcrs] = pcr_h * 300 + pcr_l;
nb_pcrs++;
if (nb_pcrs >= 2)
break;
+ } else {
+ finished_reading_packet(s, ts->raw_packet_size);
}
nb_packets++;
}
@@ -2145,15 +2156,19 @@ static int mpegts_raw_read_packet(AVFormatContext *s,
int64_t pcr_h, next_pcr_h, pos;
int pcr_l, next_pcr_l;
uint8_t pcr_buf[12];
+ uint8_t *data;
if (av_new_packet(pkt, TS_PACKET_SIZE) < 0)
return AVERROR(ENOMEM);
pkt->pos= avio_tell(s->pb);
- ret = read_packet(s, pkt->data, ts->raw_packet_size);
+ ret = read_packet(s, pkt->data, ts->raw_packet_size, &data);
if (ret < 0) {
av_free_packet(pkt);
return ret;
}
+ if (data != pkt->data)
+ memcpy(pkt->data, data, ts->raw_packet_size);
+ finished_reading_packet(s, ts->raw_packet_size);
if (ts->mpeg2ts_compute_pcr) {
/* compute exact PCR for each packet */
if (parse_pcr(&pcr_h, &pcr_l, pkt->data) == 0) {
--
1.7.9.5

47
packages/multimedia/ffmpeg/patches/ffmpeg-0053-ffmpeg-backport-mpegts-Make-discard_pid-faster-for-s.patch
View File

@ -1,47 +0,0 @@
From 6aec5772fd5331b3514f308ab0895f6234b60045 Mon Sep 17 00:00:00 2001
From: Ben Avison <bavison@riscosopen.org>
Date: Mon, 5 Aug 2013 13:12:51 +0100
Subject: [PATCH 53/54] [ffmpeg] - backport - mpegts: Make discard_pid()
faster for single-program streams
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
When a stream contains a single program, there's no point in doing a
PID -> program lookup. Normally the one and only program isn't disabled,
so no packets should be discarded.
Before After
Mean StdDev Mean StdDev Change
discard_pid() 73.8 9.4 20.2 1.5 +264.8%
Overall 2300.8 28.0 2253.1 20.6 +2.1%
Signed-off-by: Martin Storsjö <martin@martin.st>
---
libavformat/mpegts.c | 11 +++++++++++
1 file changed, 11 insertions(+)
diff --git a/libavformat/mpegts.c b/libavformat/mpegts.c
index 5307521..82dd209 100644
--- a/libavformat/mpegts.c
+++ b/libavformat/mpegts.c
@@ -268,6 +268,17 @@ static int discard_pid(MpegTSContext *ts, unsigned int pid)
int i, j, k;
int used = 0, discarded = 0;
struct Program *p;
+
+ /* If none of the programs have .discard=AVDISCARD_ALL then there's
+ * no way we have to discard this packet
+ */
+ for (k = 0; k < ts->stream->nb_programs; k++) {
+ if (ts->stream->programs[k]->discard == AVDISCARD_ALL)
+ break;
+ }
+ if (k == ts->stream->nb_programs)
+ return 0;
+
for(i=0; i<ts->nb_prg; i++) {
p = &ts->prg[i];
for(j=0; j<p->nb_pids; j++) {
--
1.7.9.5

76
packages/multimedia/ffmpeg/patches/ffmpeg-0054-ffmpeg-backport-mpegts-Remove-one-64-bit-integer-mod.patch
View File

@ -1,76 +0,0 @@
From b79aa2b89ed9027a72a10c1d26ccdf2bb385d57b Mon Sep 17 00:00:00 2001
From: Ben Avison <bavison@riscosopen.org>
Date: Mon, 5 Aug 2013 13:12:49 +0100
Subject: [PATCH 54/54] [ffmpeg] - backport - mpegts: Remove one 64-bit
integer modulus operation per packet
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
The common case of the pointer having increased by one packet (which results
in no change to the modulus) can be detected with a 64-bit subtraction,
which is far cheaper than a division on many platforms.
Before After
Mean StdDev Mean StdDev Change
Divisions 248.3 8.8 51.5 7.4 +381.7%
Overall 2773.2 25.6 2372.5 43.1 +16.9%
Signed-off-by: Martin Storsjö <martin@martin.st>
---
libavcodec/mathops.h | 9 +++++++++
libavformat/mpegts.c | 5 ++++-
2 files changed, 13 insertions(+), 1 deletion(-)
diff --git a/libavcodec/mathops.h b/libavcodec/mathops.h
index 592f5a5..1d57342 100644
--- a/libavcodec/mathops.h
+++ b/libavcodec/mathops.h
@@ -195,6 +195,15 @@ if ((y) < (x)) {\
# define FASTDIV(a,b) ((uint32_t)((((uint64_t)a) * ff_inverse[b]) >> 32))
#endif /* FASTDIV */
+#ifndef MOD_UNLIKELY
+# define MOD_UNLIKELY(modulus, dividend, divisor, prev_dividend) \
+ do { \
+ if ((prev_dividend) == 0 || (dividend) - (prev_dividend) != (divisor)) \
+ (modulus) = (dividend) % (divisor); \
+ (prev_dividend) = (dividend); \
+ } while (0)
+#endif
+
static inline av_const unsigned int ff_sqrt(unsigned int a)
{
unsigned int b;
diff --git a/libavformat/mpegts.c b/libavformat/mpegts.c
index 82dd209..b995f60 100644
--- a/libavformat/mpegts.c
+++ b/libavformat/mpegts.c
@@ -28,6 +28,7 @@
#include "libavutil/avassert.h"
#include "libavcodec/bytestream.h"
#include "libavcodec/get_bits.h"
+#include "libavcodec/mathops.h"
#include "avformat.h"
#include "mpegts.h"
#include "internal.h"
@@ -99,6 +100,8 @@ struct MpegTSContext {
int raw_packet_size;
int pos47;
+ /** position corresponding to pos47, or 0 if pos47 invalid */
+ int64_t pos;
/** if true, all pids are analyzed to find streams */
int auto_guess;
@@ -1814,7 +1817,7 @@ static int handle_packet(MpegTSContext *ts, const uint8_t *packet)
return 0;
pos = avio_tell(ts->stream->pb);
- ts->pos47= pos % ts->raw_packet_size;
+ MOD_UNLIKELY(ts->pos47, pos, ts->raw_packet_size, ts->pos);
if (tss->type == MPEGTS_SECTION) {
if (is_start) {
--
1.7.9.5

110
packages/multimedia/ffmpeg/patches/ffmpeg-0055-ffmpeg-backport-fix-compilation-droid-gcc-4-7.patch
View File

@ -1,110 +0,0 @@
From 8067f55edf3719182aed6e5b57b7863889f80218 Mon Sep 17 00:00:00 2001
From: =?utf8?q?Reimar=20D=C3=B6ffinger?= <Reimar.Doeffinger@gmx.de>
Date: Sat, 16 Mar 2013 13:36:20 +0100
Subject: [PATCH] Fix compilation on ARM with android gcc 4.7
MIME-Version: 1.0
Content-Type: text/plain; charset=utf8
Content-Transfer-Encoding: 8bit
With the current code it fails due to running out
of registers.
So code the store offsets manually into the assembler
instead.
Passes "make fate-dts".
Signed-off-by: Reimar Döffinger <Reimar.Doeffinger@gmx.de>
---
libavcodec/arm/dca.h | 74 ++++++++++++++++++++++++--------------------------
1 file changed, 36 insertions(+), 38 deletions(-)
diff --git a/libavcodec/arm/dca.h b/libavcodec/arm/dca.h
index 2cfd18a..431b62e 100644
--- a/libavcodec/arm/dca.h
+++ b/libavcodec/arm/dca.h
@@ -34,46 +34,44 @@ static inline int decode_blockcodes(int code1, int code2, int levels,
{
int v0, v1, v2, v3, v4, v5;
- __asm__ ("smmul %8, %14, %18 \n"
- "smmul %11, %15, %18 \n"
- "smlabb %14, %8, %17, %14 \n"
- "smlabb %15, %11, %17, %15 \n"
- "smmul %9, %8, %18 \n"
- "smmul %12, %11, %18 \n"
- "sub %14, %14, %16, lsr #1 \n"
- "sub %15, %15, %16, lsr #1 \n"
- "smlabb %8, %9, %17, %8 \n"
- "smlabb %11, %12, %17, %11 \n"
- "smmul %10, %9, %18 \n"
- "smmul %13, %12, %18 \n"
- "str %14, %0 \n"
- "str %15, %4 \n"
- "sub %8, %8, %16, lsr #1 \n"
- "sub %11, %11, %16, lsr #1 \n"
- "smlabb %9, %10, %17, %9 \n"
- "smlabb %12, %13, %17, %12 \n"
- "smmul %14, %10, %18 \n"
- "smmul %15, %13, %18 \n"
- "str %8, %1 \n"
- "str %11, %5 \n"
- "sub %9, %9, %16, lsr #1 \n"
- "sub %12, %12, %16, lsr #1 \n"
- "smlabb %10, %14, %17, %10 \n"
- "smlabb %13, %15, %17, %13 \n"
- "str %9, %2 \n"
- "str %12, %6 \n"
- "sub %10, %10, %16, lsr #1 \n"
- "sub %13, %13, %16, lsr #1 \n"
- "str %10, %3 \n"
- "str %13, %7 \n"
- : "=m"(values[0]), "=m"(values[1]),
- "=m"(values[2]), "=m"(values[3]),
- "=m"(values[4]), "=m"(values[5]),
- "=m"(values[6]), "=m"(values[7]),
- "=&r"(v0), "=&r"(v1), "=&r"(v2),
+ __asm__ ("smmul %0, %6, %10 \n"
+ "smmul %3, %7, %10 \n"
+ "smlabb %6, %0, %9, %6 \n"
+ "smlabb %7, %3, %9, %7 \n"
+ "smmul %1, %0, %10 \n"
+ "smmul %4, %3, %10 \n"
+ "sub %6, %6, %8, lsr #1 \n"
+ "sub %7, %7, %8, lsr #1 \n"
+ "smlabb %0, %1, %9, %0 \n"
+ "smlabb %3, %4, %9, %3 \n"
+ "smmul %2, %1, %10 \n"
+ "smmul %5, %4, %10 \n"
+ "str %6, [%11, #0] \n"
+ "str %7, [%11, #16] \n"
+ "sub %0, %0, %8, lsr #1 \n"
+ "sub %3, %3, %8, lsr #1 \n"
+ "smlabb %1, %2, %9, %1 \n"
+ "smlabb %4, %5, %9, %4 \n"
+ "smmul %6, %2, %10 \n"
+ "smmul %7, %5, %10 \n"
+ "str %0, [%11, #4] \n"
+ "str %3, [%11, #20] \n"
+ "sub %1, %1, %8, lsr #1 \n"
+ "sub %4, %4, %8, lsr #1 \n"
+ "smlabb %2, %6, %9, %2 \n"
+ "smlabb %5, %7, %9, %5 \n"
+ "str %1, [%11, #8] \n"
+ "str %4, [%11, #24] \n"
+ "sub %2, %2, %8, lsr #1 \n"
+ "sub %5, %5, %8, lsr #1 \n"
+ "str %2, [%11, #12] \n"
+ "str %5, [%11, #28] \n"
+ : "=&r"(v0), "=&r"(v1), "=&r"(v2),
"=&r"(v3), "=&r"(v4), "=&r"(v5),
"+&r"(code1), "+&r"(code2)
- : "r"(levels - 1), "r"(-levels), "r"(ff_inverse[levels]));
+ : "r"(levels - 1), "r"(-levels),
+ "r"(ff_inverse[levels]), "r"(values)
+ : "memory");
return code1 | code2;
}
--
1.7.10.4

14
packages/multimedia/ffmpeg/patches/ffmpeg-0056-backport-h264-do-not-discard-NAL_SEI-when-skipping-frames.patch
View File

@ -1,14 +0,0 @@
diff --git a/libavcodec/h264.c b/libavcodec/h264.c
index 85f2372..1cab49f 100644
--- a/libavcodec/h264.c
+++ b/libavcodec/h264.c
@@ -4523,8 +4523,7 @@ static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size,
first_slice = hx->nal_unit_type;
}
- // FIXME do not discard SEI id
- if (avctx->skip_frame >= AVDISCARD_NONREF && h->nal_ref_idc == 0)
+ if (avctx->skip_frame >= AVDISCARD_NONREF && h->nal_ref_idc == 0 && h->nal_unit_type != NAL_SEI)
continue;
again:

84
packages/multimedia/ffmpeg/patches/ffmpeg-0057-backport-swresample-Make-rematrix-maxvalue-user-settable.patch
View File

@ -1,84 +0,0 @@
From e2b718464e92fcde3d21c6653c88ddec2ab21c3f Mon Sep 17 00:00:00 2001
From: Michael Niedermayer <michaelni@gmx.at>
Date: Mon, 22 Jul 2013 03:23:54 +0200
Subject: [PATCH] swresample: Make rematrix maxvalue user settable
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
---
libswresample/rematrix.c | 13 +++++++++++--
libswresample/swresample.c | 1 +
libswresample/swresample_internal.h | 1 +
libswresample/version.h | 2 +-
4 files changed, 14 insertions(+), 3 deletions(-)
diff --git a/libswresample/rematrix.c b/libswresample/rematrix.c
index 48aff3b..33e2966 100644
--- a/libswresample/rematrix.c
+++ b/libswresample/rematrix.c
@@ -120,6 +120,7 @@ av_cold static int auto_matrix(SwrContext *s)
double maxcoef=0;
char buf[128];
const int matrix_encoding = s->matrix_encoding;
+ float maxval;
in_ch_layout = clean_layout(s, s->in_ch_layout);
if(!sane_layout(in_ch_layout)){
@@ -304,8 +305,16 @@ av_cold static int auto_matrix(SwrContext *s)
if(s->rematrix_volume < 0)
maxcoef = -s->rematrix_volume;
- if(( av_get_packed_sample_fmt(s->out_sample_fmt) < AV_SAMPLE_FMT_FLT
- || av_get_packed_sample_fmt(s->int_sample_fmt) < AV_SAMPLE_FMT_FLT) && maxcoef > 1.0){
+ if (s->rematrix_maxval > 0) {
+ maxval = s->rematrix_maxval;
+ } else if ( av_get_packed_sample_fmt(s->out_sample_fmt) < AV_SAMPLE_FMT_FLT
+ || av_get_packed_sample_fmt(s->int_sample_fmt) < AV_SAMPLE_FMT_FLT) {
+ maxval = 1.0;
+ } else
+ maxval = INT_MAX;
+
+ if(maxcoef > maxval){
+ maxcoef /= maxval;
for(i=0; i<SWR_CH_MAX; i++)
for(j=0; j<SWR_CH_MAX; j++){
s->matrix[i][j] /= maxcoef;
diff --git a/libswresample/swresample.c b/libswresample/swresample.c
index ba2afdb..cdfe5bf 100644
--- a/libswresample/swresample.c
+++ b/libswresample/swresample.c
@@ -68,6 +68,7 @@
{"lfe_mix_level" , "set LFE mix level" , OFFSET(lfe_mix_level ), AV_OPT_TYPE_FLOAT, {.dbl=0 }, -32 , 32 , PARAM},
{"rmvol" , "set rematrix volume" , OFFSET(rematrix_volume), AV_OPT_TYPE_FLOAT, {.dbl=1.0 }, -1000 , 1000 , PARAM},
{"rematrix_volume" , "set rematrix volume" , OFFSET(rematrix_volume), AV_OPT_TYPE_FLOAT, {.dbl=1.0 }, -1000 , 1000 , PARAM},
+{"rematrix_maxval" , "set rematrix maxval" , OFFSET(rematrix_maxval), AV_OPT_TYPE_FLOAT, {.dbl=0.0 }, 0 , 1000 , PARAM},
{"flags" , "set flags" , OFFSET(flags ), AV_OPT_TYPE_FLAGS, {.i64=0 }, 0 , UINT_MAX , PARAM, "flags"},
{"swr_flags" , "set flags" , OFFSET(flags ), AV_OPT_TYPE_FLAGS, {.i64=0 }, 0 , UINT_MAX , PARAM, "flags"},
diff --git a/libswresample/swresample_internal.h b/libswresample/swresample_internal.h
index 108f837..ab19f21 100644
--- a/libswresample/swresample_internal.h
+++ b/libswresample/swresample_internal.h
@@ -82,6 +82,7 @@ struct SwrContext {
float clev; ///< center mixing level
float lfe_mix_level; ///< LFE mixing level
float rematrix_volume; ///< rematrixing volume coefficient
+ float rematrix_maxval; ///< maximum value for rematrixing output
enum AVMatrixEncoding matrix_encoding; /**< matrixed stereo encoding */
const int *channel_map; ///< channel index (or -1 if muted channel) map
int used_ch_count; ///< number of used input channels (mapped channel count if channel_map, otherwise in.ch_count)
diff --git a/libswresample/version.h b/libswresample/version.h
index df9df48..8272b76 100644
--- a/libswresample/version.h
+++ b/libswresample/version.h
@@ -30,7 +30,7 @@
#define LIBSWRESAMPLE_VERSION_MAJOR 0
#define LIBSWRESAMPLE_VERSION_MINOR 17
-#define LIBSWRESAMPLE_VERSION_MICRO 102
+#define LIBSWRESAMPLE_VERSION_MICRO 103
#define LIBSWRESAMPLE_VERSION_INT AV_VERSION_INT(LIBSWRESAMPLE_VERSION_MAJOR, \
LIBSWRESAMPLE_VERSION_MINOR, \
--
1.8.4

53
packages/multimedia/ffmpeg/patches/ffmpeg-0058-backport-fixed-not-keeping-custom-http-headers.patch
View File

@ -1,53 +0,0 @@
From 9fb6300b1cc36c8111cee017fbb8a3e37b0c4e68 Mon Sep 17 00:00:00 2001
From: siriuzwhite <siriuz@gmx.net>
Date: Sun, 10 Nov 2013 23:26:16 +0100
Subject: [PATCH] [backport] fixed ffmpeg not keep custom http headers when
playing hls stream
---
libavformat/hls.c | 9 +++++++++
1 file changed, 9 insertions(+)
diff --git a/libavformat/hls.c b/libavformat/hls.c
index 7de6059..2e3fd00 100644
--- a/libavformat/hls.c
+++ b/libavformat/hls.c
@@ -105,6 +105,7 @@ struct variant {
AVIOInterruptCB *interrupt_callback;
char *user_agent; ///< holds HTTP user agent set as an AVOption to the HTTP protocol context
char *cookies; ///< holds HTTP cookie values set in either the initial response or as an AVOption to the HTTP protocol context
+ char *headers; ///< holds HTTP headers set as an AVOption to the HTTP protocol context (kow|mitm)
} HLSContext;
static int read_chomp_line(AVIOContext *s, char *buf, int maxlen)
@@ -224,6 +225,7 @@ static int parse_playlist(HLSContext *c, const char *url,
// broker prior HTTP options that should be consistent across requests
av_dict_set(&opts, "user-agent", c->user_agent, 0);
av_dict_set(&opts, "cookies", c->cookies, 0);
+ av_dict_set(&opts, "headers", c->headers, 0); // (kow|mitm)
ret = avio_open2(&in, url, AVIO_FLAG_READ,
c->interrupt_callback, &opts);
@@ -346,6 +348,7 @@ static int open_input(HLSContext *c, struct variant *var)
// broker prior HTTP options that should be consistent across requests
av_dict_set(&opts, "user-agent", c->user_agent, 0);
av_dict_set(&opts, "cookies", c->cookies, 0);
+ av_dict_set(&opts, "headers", c->headers, 0); // (kow|mitm)
av_dict_set(&opts, "seekable", "0", 0);
if (seg->key_type == KEY_NONE) {
@@ -494,6 +497,12 @@ static int hls_read_header(AVFormatContext *s)
av_opt_get(u->priv_data, "cookies", 0, (uint8_t**)&(c->cookies));
if (c->cookies && !strlen(c->cookies))
av_freep(&c->cookies);
+
+ // get the previous headers & set back to null if string size is zero (kow|mitm)
+ av_freep(&c->headers);
+ av_opt_get(u->priv_data, "headers", 0, (uint8_t**)&(c->headers));
+ if (c->headers && !strlen(c->headers))
+ av_freep(&c->headers);
}
if ((ret = parse_playlist(c, s->filename, NULL, s->pb)) < 0)
--
1.8.4

29
packages/multimedia/ffmpeg/patches/ffmpeg-0059-backport-mpegts-do-not-set-pts-for-missing-dts.patch
View File

@ -1,29 +0,0 @@
From f65afef1df49f53e14c8d4173ff960fff8d44ecb Mon Sep 17 00:00:00 2001
From: Rainer Hochecker <fernetmenta@online.de>
Date: Tue, 3 Dec 2013 10:03:04 +0100
Subject: [PATCH] mpegts: do not set pts for missing dts in video streams
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
---
libavformat/mpegts.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/libavformat/mpegts.c b/libavformat/mpegts.c
index 59b0058..d67c63a 100644
--- a/libavformat/mpegts.c
+++ b/libavformat/mpegts.c
@@ -972,7 +972,10 @@ static int mpegts_push_data(MpegTSFilter *filter,
pes->pts = AV_NOPTS_VALUE;
pes->dts = AV_NOPTS_VALUE;
if ((flags & 0xc0) == 0x80) {
- pes->dts = pes->pts = ff_parse_pes_pts(r);
+ pes->pts = ff_parse_pes_pts(r);
+ /* video pts is not monotonic, can't be used for dts */
+ if (pes->st->codec->codec_type != AVMEDIA_TYPE_VIDEO)
+ pes->dts = pes->pts;
r += 5;
} else if ((flags & 0xc0) == 0xc0) {
pes->pts = ff_parse_pes_pts(r);
--
1.8.5-rc3

135
packages/multimedia/ffmpeg/patches/ffmpeg-0060-mpegts-stop-analyzing-when-pmt-for-all-programs-have.patch
View File

@ -1,135 +0,0 @@
From 120e44f817d73572d14a3db4af24a3d517aaacd8 Mon Sep 17 00:00:00 2001
From: Joakim Plate <elupus@ecce.se>
Date: Sat, 14 Dec 2013 14:55:13 +0100
Subject: [PATCH 1/1] mpegts: stop analyzing when pmt for all programs have
been found
This disables NOHEADER after finding PMT for all programs to
avoid find_stream_info always exhausting probe size for mpegts.
This is very important for live streams since read speed
will be limited. rtsp, udp and any protocol streaming a live
mpegts will have dramatically faster startup time.
Note, lack of codec parameters for streams can still cause
the full probe size to be exhausted.
---
libavformat/mpegts.c | 54 +++++++++++++++++++++++++++++++++++++++++++---------
1 file changed, 45 insertions(+), 9 deletions(-)
diff --git a/libavformat/mpegts.c b/libavformat/mpegts.c
index d67c63a..fa92fb7 100644
--- a/libavformat/mpegts.c
+++ b/libavformat/mpegts.c
@@ -90,6 +90,9 @@ struct Program {
unsigned int id; //program id/service id
unsigned int nb_pids;
unsigned int pids[MAX_PIDS_PER_PROGRAM];
+
+ /** have we found pmt for this program */
+ int pmt_found;
};
struct MpegTSContext {
@@ -205,6 +208,17 @@ typedef struct PESContext {
extern AVInputFormat ff_mpegts_demuxer;
+static struct Program * get_program(MpegTSContext *ts, unsigned int programid)
+{
+ int i;
+ for(i=0; i<ts->nb_prg; i++) {
+ if(ts->prg[i].id == programid) {
+ return &ts->prg[i];
+ }
+ }
+ return NULL;
+}
+
static void clear_avprogram(MpegTSContext *ts, unsigned int programid)
{
AVProgram *prg = NULL;
@@ -225,8 +239,10 @@ static void clear_program(MpegTSContext *ts, unsigned int programid)
clear_avprogram(ts, programid);
for(i=0; i<ts->nb_prg; i++)
- if(ts->prg[i].id == programid)
+ if(ts->prg[i].id == programid) {
ts->prg[i].nb_pids = 0;
+ ts->prg[i].pmt_found = 0;
+ }
}
static void clear_programs(MpegTSContext *ts)
@@ -245,19 +261,13 @@ static void add_pat_entry(MpegTSContext *ts, unsigned int programid)
p = &ts->prg[ts->nb_prg];
p->id = programid;
p->nb_pids = 0;
+ p->pmt_found = 0;
ts->nb_prg++;
}
static void add_pid_to_pmt(MpegTSContext *ts, unsigned int programid, unsigned int pid)
{
- int i;
- struct Program *p = NULL;
- for(i=0; i<ts->nb_prg; i++) {
- if(ts->prg[i].id == programid) {
- p = &ts->prg[i];
- break;
- }
- }
+ struct Program *p = get_program(ts, programid);
if(!p)
return;
@@ -266,6 +276,15 @@ static void add_pid_to_pmt(MpegTSContext *ts, unsigned int programid, unsigned i
p->pids[p->nb_pids++] = pid;
}
+static void set_pmt_found(MpegTSContext *ts, unsigned int programid)
+{
+ struct Program *p = get_program(ts, programid);
+ if(!p)
+ return;
+
+ p->pmt_found = 1;
+}
+
static void set_pcr_pid(AVFormatContext *s, unsigned int programid, unsigned int pid)
{
int i;
@@ -1590,6 +1609,8 @@ static void pmt_cb(MpegTSFilter *filter, const uint8_t *section, int section_len
if (!ts->stream->nb_streams)
ts->stop_parse = 2;
+ set_pmt_found(ts, h->id);
+
for(;;) {
st = 0;
pes = NULL;
@@ -1912,6 +1933,21 @@ static int handle_packet(MpegTSContext *ts, const uint8_t *packet)
p, p_end - p, 0);
}
}
+
+ // stop find_stream_info from waiting for more streams
+ // when all programs have received a PMT
+ if( ts->stream->ctx_flags & AVFMTCTX_NOHEADER) {
+ int i;
+ for(i=0; i<ts->nb_prg; i++) {
+ if (!ts->prg[i].pmt_found)
+ break;
+ }
+ if (i == ts->nb_prg && ts->nb_prg > 0) {
+ av_log(ts->stream, AV_LOG_DEBUG, "All programs have pmt, headers found\n");
+ ts->stream->ctx_flags &= ~AVFMTCTX_NOHEADER;
+ }
+ }
+
} else {
int ret;
int64_t pcr = -1;
--
1.8.2

29
packages/multimedia/ffmpeg/patches/ffmpeg-0061-ffmpeg-backport-Correctly-handle-cookies-that-specify-a-sub-domain.patch
View File

@ -1,29 +0,0 @@
From aa1852f88f769ee438e85cec8552ff545914885c Mon Sep 17 00:00:00 2001
From: Eli K. <eli@algotec.co.il>
Date: Thu, 23 Jan 2014 10:27:21 +0200
Subject: [PATCH] Correctly handle cookies that specify a sub-domain.
---
libavformat/http.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/libavformat/http.c b/libavformat/http.c
index 3b655c6..69c4d6d 100644
--- a/libavformat/http.c
+++ b/libavformat/http.c
@@ -490,8 +490,11 @@ static int get_cookies(HTTPContext *s, char **cookies, const char *path,
av_free(cpath);
cpath = av_strdup(&param[5]);
} else if (!av_strncasecmp("domain=", param, 7)) {
+ // if the cookie specifies a sub-domain, skip the leading dot thereby
+ // supporting URLs that point to sub-domains and the master domain
+ int leading_dot = (param[7] == '.');
av_free(cdomain);
- cdomain = av_strdup(&param[7]);
+ cdomain = av_strdup(&param[7+leading_dot]);
} else if (!av_strncasecmp("secure", param, 6) ||
!av_strncasecmp("comment", param, 7) ||
!av_strncasecmp("max-age", param, 7) ||
--
1.8.3.msysgit.0

67
packages/multimedia/ffmpeg/patches/ffmpeg-0061-ffmpeg-backport-avcodec-aacdec-default-to-non-wide-7.patch
View File

@ -1,67 +0,0 @@
From 59f16d529fdcb4c5db53ce3bd289aa8148a13de4 Mon Sep 17 00:00:00 2001
From: Anssi Hannula <anssi.hannula@iki.fi>
Date: Tue, 17 Dec 2013 23:04:31 +0200
Subject: [PATCH] [ffmpeg] - backport - avcodec/aacdec: default to non-wide 7.1
in non-strict mode
Upstream commit e10fccf62a36e09b54ad6ea3d5fa6638f298d5ae, for
http://trac.xbmc.org/ticket/13758.
AAC specification has 7.1(wide) as a default layout for 8-channel
streams (channel config 7). However, at least Nero AAC encoder encodes
non-wide 7.1 streams using the default channel config 7, mapping the
side channels of the original audio stream to the second
AAC_CHANNEL_FRONT pair in the AAC stream. Similarly, e.g. FAAD decodes
the second AAC_CHANNEL_FRONT pair as side channels, therefore decoding
the incorrect streams as if they were correct (and as the encoder
intended).
FFmpeg currently decodes such files by-the-spec, i.e. after decoding the
original front pair will be in AV_CH_FRONT_x_OF_CENTER and the original
side pair will be in AV_CH_FRONT_x.
As actual intended 7.1(wide) streams are very rare while misencoded 7.1
files actually exist in the wild, default to assuming a 7.1 layout was
intended unless in strict mode.
Fixes playback of e.g. 8_Channel_ID.m4a in samples.
Signed-off-by: Anssi Hannula <anssi.hannula@iki.fi>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
---
libavcodec/aacdec.c | 19 +++++++++++++++++++
1 file changed, 19 insertions(+)
diff --git a/libavcodec/aacdec.c b/libavcodec/aacdec.c
index 7a871c4..12dbfcf 100644
--- a/libavcodec/aacdec.c
+++ b/libavcodec/aacdec.c
@@ -505,6 +505,25 @@ static int set_default_channel_config(AVCodecContext *avctx,
}
*tags = tags_per_config[channel_config];
memcpy(layout_map, aac_channel_layout_map[channel_config-1], *tags * sizeof(*layout_map));
+
+ /*
+ * AAC specification has 7.1(wide) as a default layout for 8-channel streams.
+ * However, at least Nero AAC encoder encodes 7.1 streams using the default
+ * channel config 7, mapping the side channels of the original audio stream
+ * to the second AAC_CHANNEL_FRONT pair in the AAC stream. Similarly, e.g. FAAD
+ * decodes the second AAC_CHANNEL_FRONT pair as side channels, therefore decoding
+ * the incorrect streams as if they were correct (and as the encoder intended).
+ *
+ * As actual intended 7.1(wide) streams are very rare, default to assuming a
+ * 7.1 layout was intended.
+ */
+ if (channel_config == 7 && avctx->strict_std_compliance < FF_COMPLIANCE_STRICT) {
+ av_log(avctx, AV_LOG_INFO, "Assuming an incorrectly encoded 7.1 channel layout"
+ " instead of a spec-compliant 7.1(wide) layout, use -strict %d to decode"
+ " according to the specification instead.\n", FF_COMPLIANCE_STRICT);
+ layout_map[2][2] = AAC_CHANNEL_SIDE;
+ }
+
return 0;
}
--
1.8.1.5

40
packages/multimedia/ffmpeg/patches/ffmpeg-999-0001-h264_parser-Initialize-the-h264dsp-context-in-the-.patch
View File

@ -1,40 +0,0 @@
From 6f99f0779ea56e4bfe40f7ca56e60b3dfd84eba6 Mon Sep 17 00:00:00 2001
From: Ben Avison <bavison@riscosopen.org>
Date: Mon, 5 Aug 2013 13:12:46 +0100
Subject: [PATCH] h264_parser: Initialize the h264dsp context in the parser as
well
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Each AVStream struct for an H.264 elementary stream actually has two
copies of the H264DSPContext struct (and in fact all the other members
of H264Context as well):
((H264Context *) ((AVStream *)st)->codec->priv_data)->h264dsp
((H264Context *) ((AVStream *)st)->parser->priv_data)->h264dsp
but only the first of these was actually being initialised. This
prevented the addition of platform-specific implementations of
parser-related functions.
Signed-off-by: Martin Storsjö <martin@martin.st>
---
libavcodec/h264_parser.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/libavcodec/h264_parser.c b/libavcodec/h264_parser.c
index aff9ba1..a732f79 100644
--- a/libavcodec/h264_parser.c
+++ b/libavcodec/h264_parser.c
@@ -386,6 +386,7 @@ static int init(AVCodecParserContext *s)
H264Context *h = s->priv_data;
h->thread_context[0] = h;
h->slice_context_count = 1;
+ ff_h264dsp_init(&h->h264dsp, 8, 1);
return 0;
}
--
1.8.5.1

128
packages/multimedia/ffmpeg/patches/ffmpeg-999-0002-h264dsp-Factorize-code-into-a-new-function-.patch
View File

@ -1,128 +0,0 @@
From 971a57f6067c96f8dba087285065618f1ac3ecd5 Mon Sep 17 00:00:00 2001
From: Ben Avison <bavison@riscosopen.org>
Date: Mon, 5 Aug 2013 13:12:47 +0100
Subject: [PATCH] h264dsp: Factorize code into a new function,
h264_find_start_code_candidate
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
This performs the start code search which was previously part of
h264_find_frame_end() - the most CPU intensive part of the function.
By itself, this results in a performance regression:
Before After
Mean StdDev Mean StdDev Change
Overall time 2925.6 26.2 3068.5 31.7 -4.7%
but this can more than be made up for by platform-optimised
implementations of the function.
Signed-off-by: Martin Storsjö <martin@martin.st>
---
libavcodec/h264_parser.c | 20 +++-----------------
libavcodec/h264dsp.c | 29 +++++++++++++++++++++++++++++
libavcodec/h264dsp.h | 9 +++++++++
3 files changed, 41 insertions(+), 17 deletions(-)
diff --git a/libavcodec/h264_parser.c b/libavcodec/h264_parser.c
index a732f79..972aace 100644
--- a/libavcodec/h264_parser.c
+++ b/libavcodec/h264_parser.c
@@ -62,23 +62,9 @@ static int ff_h264_find_frame_end(H264Context *h, const uint8_t *buf, int buf_si
}
if(state==7){
-#if HAVE_FAST_UNALIGNED
- /* we check i<buf_size instead of i+3/7 because its simpler
- * and there should be FF_INPUT_BUFFER_PADDING_SIZE bytes at the end
- */
-# if HAVE_FAST_64BIT
- while(i<next_avc && !((~*(const uint64_t*)(buf+i) & (*(const uint64_t*)(buf+i) - 0x0101010101010101ULL)) & 0x8080808080808080ULL))
- i+=8;
-# else
- while(i<next_avc && !((~*(const uint32_t*)(buf+i) & (*(const uint32_t*)(buf+i) - 0x01010101U)) & 0x80808080U))
- i+=4;
-# endif
-#endif
- for(; i<next_avc; i++){
- if(!buf[i]){
- state=2;
- break;
- }
+ i += h->h264dsp.h264_find_start_code_candidate(buf + i, buf_size - i);
+ if (i < buf_size)
+ state = 2;
}
}else if(state<=2){
if(buf[i]==1) state^= 5; //2->7, 1->4, 0->5
diff --git a/libavcodec/h264dsp.c b/libavcodec/h264dsp.c
index da9e417..b7d61cd 100644
--- a/libavcodec/h264dsp.c
+++ b/libavcodec/h264dsp.c
@@ -60,6 +60,34 @@
#include "h264addpx_template.c"
#undef BIT_DEPTH
+static int h264_find_start_code_candidate_c(const uint8_t *buf, int size)
+{
+ int i = 0;
+#if HAVE_FAST_UNALIGNED
+ /* we check i < size instead of i + 3 / 7 because it is
+ * simpler and there must be FF_INPUT_BUFFER_PADDING_SIZE
+ * bytes at the end.
+ */
+#if HAVE_FAST_64BIT
+ while (i < size &&
+ !((~*(const uint64_t *)(buf + i) &
+ (*(const uint64_t *)(buf + i) - 0x0101010101010101ULL)) &
+ 0x8080808080808080ULL))
+ i += 8;
+#else
+ while (i < size &&
+ !((~*(const uint32_t *)(buf + i) &
+ (*(const uint32_t *)(buf + i) - 0x01010101U)) &
+ 0x80808080U))
+ i += 4;
+#endif
+#endif
+ for (; i < size; i++)
+ if (!buf[i])
+ break;
+ return i;
+}
+
void ff_h264dsp_init(H264DSPContext *c, const int bit_depth, const int chroma_format_idc)
{
#undef FUNC
@@ -146,6 +174,7 @@ void ff_h264dsp_init(H264DSPContext *c, const int bit_depth, const int chroma_fo
H264_DSP(8);
break;
}
+ c->h264_find_start_code_candidate = h264_find_start_code_candidate_c;
if (ARCH_ARM) ff_h264dsp_init_arm(c, bit_depth, chroma_format_idc);
if (HAVE_ALTIVEC) ff_h264dsp_init_ppc(c, bit_depth, chroma_format_idc);
diff --git a/libavcodec/h264dsp.h b/libavcodec/h264dsp.h
index 98ea15c..1be4804 100644
--- a/libavcodec/h264dsp.h
+++ b/libavcodec/h264dsp.h
@@ -105,6 +105,15 @@ typedef void (*h264_biweight_func)(uint8_t *dst, uint8_t *src,
/* bypass-transform */
void (*h264_add_pixels8_clear)(uint8_t *dst, int16_t *block, int stride);
void (*h264_add_pixels4_clear)(uint8_t *dst, int16_t *block, int stride);
+
+ /**
+ * Search buf from the start for up to size bytes. Return the index
+ * of a zero byte, or >= size if not found. Ideally, use lookahead
+ * to filter out any zero bytes that are known to not be followed by
+ * one or more further zero bytes and a one byte. Better still, filter
+ * out any bytes that form the trailing_zero_8bits syntax element too.
+ */
+ int (*h264_find_start_code_candidate)(const uint8_t *buf, int size);
} H264DSPContext;
void ff_h264dsp_init(H264DSPContext *c, const int bit_depth,
--
1.8.5.1

335
packages/multimedia/ffmpeg/patches/ffmpeg-999-0003-arm-Add-assembly-version-of-h264_find_start_code_candidate.patch
View File

@ -1,335 +0,0 @@
From fdc814cc6701f3e882a7ea7f29d16500c7340f0d Mon Sep 17 00:00:00 2001
From: Ben Avison <bavison@riscosopen.org>
Date: Mon, 5 Aug 2013 13:12:48 +0100
Subject: [PATCH] arm: Add assembly version of h264_find_start_code_candidate
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Before After
Mean StdDev Mean StdDev Change
This function 508.8 23.4 185.4 9.0 +174.4%
Overall 3068.5 31.7 2752.1 29.4 +11.5%
In combination with the preceding patch:
Before After
Mean StdDev Mean StdDev Change
Overall 2925.6 26.2 2752.1 29.4 +6.3%
Signed-off-by: Martin Storsjö <martin@martin.st>
---
libavcodec/arm/Makefile | 1 +
libavcodec/arm/h264dsp_armv6.S | 253 +++++++++++++++++++++++++++
libavcodec/arm/h264dsp_init_arm.c | 4 +
libavcodec/h264_parser.c | 1 -
4 files changed, 258 insertions(+), 1 deletion(-)
create mode 100644 libavcodec/arm/h264dsp_armv6.S
diff --git a/libavcodec/arm/Makefile b/libavcodec/arm/Makefile
index 7390a8b..480000b 100644
--- a/libavcodec/arm/Makefile
+++ b/libavcodec/arm/Makefile
@@ -9,6 +9,7 @@ OBJS-$(CONFIG_AAC_DECODER) += arm/sbrdsp_init_arm.o \
OBJS-$(CONFIG_DCA_DECODER) += arm/dcadsp_init_arm.o \
ARMV6-OBJS-$(CONFIG_AC3DSP) += arm/ac3dsp_armv6.o
+ARMV6-OBJS-$(CONFIG_H264DSP) += arm/h264dsp_armv6.o
OBJS-$(CONFIG_FLAC_DECODER) += arm/flacdsp_init_arm.o \
arm/flacdsp_arm.o \
diff --git a/libavcodec/arm/h264dsp_armv6.S b/libavcodec/arm/h264dsp_armv6.S
new file mode 100644
index 0000000..c4f12a6
--- /dev/null
+++ b/libavcodec/arm/h264dsp_armv6.S
@@ -0,0 +1,253 @@
+/*
+ * Copyright (c) 2013 RISC OS Open Ltd
+ * Author: Ben Avison <bavison@riscosopen.org>
+ *
+ * This file is part of Libav.
+ *
+ * Libav 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.
+ *
+ * Libav 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 Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/arm/asm.S"
+
+RESULT .req a1
+BUF .req a1
+SIZE .req a2
+PATTERN .req a3
+PTR .req a4
+DAT0 .req v1
+DAT1 .req v2
+DAT2 .req v3
+DAT3 .req v4
+TMP0 .req v5
+TMP1 .req v6
+TMP2 .req ip
+TMP3 .req lr
+
+#define PRELOAD_DISTANCE 4
+
+.macro innerloop4
+ ldr DAT0, [PTR], #4
+ subs SIZE, SIZE, #4 @ C flag survives rest of macro
+ sub TMP0, DAT0, PATTERN, lsr #14
+ bic TMP0, TMP0, DAT0
+ ands TMP0, TMP0, PATTERN
+.endm
+
+.macro innerloop16 decrement, do_preload
+ ldmia PTR!, {DAT0,DAT1,DAT2,DAT3}
+ .ifnc "\do_preload",""
+ pld [PTR, #PRELOAD_DISTANCE*32]
+ .endif
+ .ifnc "\decrement",""
+ subs SIZE, SIZE, #\decrement @ C flag survives rest of macro
+ .endif
+ sub TMP0, DAT0, PATTERN, lsr #14
+ sub TMP1, DAT1, PATTERN, lsr #14
+ bic TMP0, TMP0, DAT0
+ bic TMP1, TMP1, DAT1
+ sub TMP2, DAT2, PATTERN, lsr #14
+ sub TMP3, DAT3, PATTERN, lsr #14
+ ands TMP0, TMP0, PATTERN
+ bic TMP2, TMP2, DAT2
+ it eq
+ andseq TMP1, TMP1, PATTERN
+ bic TMP3, TMP3, DAT3
+ itt eq
+ andseq TMP2, TMP2, PATTERN
+ andseq TMP3, TMP3, PATTERN
+.endm
+
+/* int ff_h264_find_start_code_candidate_armv6(const uint8_t *buf, int size) */
+function ff_h264_find_start_code_candidate_armv6, export=1
+ push {v1-v6,lr}
+ mov PTR, BUF
+ @ Ensure there are at least (PRELOAD_DISTANCE+2) complete cachelines to go
+ @ before using code that does preloads
+ cmp SIZE, #(PRELOAD_DISTANCE+3)*32 - 1
+ blo 60f
+
+ @ Get to word-alignment, 1 byte at a time
+ tst PTR, #3
+ beq 2f
+1: ldrb DAT0, [PTR], #1
+ sub SIZE, SIZE, #1
+ teq DAT0, #0
+ beq 90f
+ tst PTR, #3
+ bne 1b
+2: @ Get to 4-word alignment, 1 word at a time
+ ldr PATTERN, =0x80008000
+ setend be
+ tst PTR, #12
+ beq 4f
+3: innerloop4
+ bne 91f
+ tst PTR, #12
+ bne 3b
+4: @ Get to cacheline (8-word) alignment
+ tst PTR, #16
+ beq 5f
+ innerloop16 16
+ bne 93f
+5: @ Check complete cachelines, with preloading
+ @ We need to stop when there are still (PRELOAD_DISTANCE+1)
+ @ complete cachelines to go
+ sub SIZE, SIZE, #(PRELOAD_DISTANCE+2)*32
+6: innerloop16 , do_preload
+ bne 93f
+ innerloop16 32
+ bne 93f
+ bcs 6b
+ @ Preload trailing part-cacheline, if any
+ tst SIZE, #31
+ beq 7f
+ pld [PTR, #(PRELOAD_DISTANCE+1)*32]
+ @ Check remaining data without doing any more preloads. First
+ @ do in chunks of 4 words:
+7: adds SIZE, SIZE, #(PRELOAD_DISTANCE+2)*32 - 16
+ bmi 9f
+8: innerloop16 16
+ bne 93f
+ bcs 8b
+ @ Then in words:
+9: adds SIZE, SIZE, #16 - 4
+ bmi 11f
+10: innerloop4
+ bne 91f
+ bcs 10b
+11: setend le
+ @ Check second byte of final halfword
+ ldrb DAT0, [PTR, #-1]
+ teq DAT0, #0
+ beq 90f
+ @ Check any remaining bytes
+ tst SIZE, #3
+ beq 13f
+12: ldrb DAT0, [PTR], #1
+ sub SIZE, SIZE, #1
+ teq DAT0, #0
+ beq 90f
+ tst SIZE, #3
+ bne 12b
+ @ No candidate found
+13: sub RESULT, PTR, BUF
+ b 99f
+
+60: @ Small buffer - simply check by looping over bytes
+ subs SIZE, SIZE, #1
+ bcc 99f
+61: ldrb DAT0, [PTR], #1
+ subs SIZE, SIZE, #1
+ teq DAT0, #0
+ beq 90f
+ bcs 61b
+ @ No candidate found
+ sub RESULT, PTR, BUF
+ b 99f
+
+90: @ Found a candidate at the preceding byte
+ sub RESULT, PTR, BUF
+ sub RESULT, RESULT, #1
+ b 99f
+
+91: @ Found a candidate somewhere in the preceding 4 bytes
+ sub RESULT, PTR, BUF
+ sub RESULT, RESULT, #4
+ sub TMP0, DAT0, #0x20000
+ bics TMP0, TMP0, DAT0
+ itt pl
+ ldrbpl DAT0, [PTR, #-3]
+ addpl RESULT, RESULT, #2
+ bpl 92f
+ teq RESULT, #0
+ beq 98f @ don't look back a byte if found at first byte in buffer
+ ldrb DAT0, [PTR, #-5]
+92: teq DAT0, #0
+ it eq
+ subeq RESULT, RESULT, #1
+ b 98f
+
+93: @ Found a candidate somewhere in the preceding 16 bytes
+ sub RESULT, PTR, BUF
+ sub RESULT, RESULT, #16
+ teq TMP0, #0
+ beq 95f @ not in first 4 bytes
+ sub TMP0, DAT0, #0x20000
+ bics TMP0, TMP0, DAT0
+ itt pl
+ ldrbpl DAT0, [PTR, #-15]
+ addpl RESULT, RESULT, #2
+ bpl 94f
+ teq RESULT, #0
+ beq 98f @ don't look back a byte if found at first byte in buffer
+ ldrb DAT0, [PTR, #-17]
+94: teq DAT0, #0
+ it eq
+ subeq RESULT, RESULT, #1
+ b 98f
+95: add RESULT, RESULT, #4
+ teq TMP1, #0
+ beq 96f @ not in next 4 bytes
+ sub TMP1, DAT1, #0x20000
+ bics TMP1, TMP1, DAT1
+ itee mi
+ ldrbmi DAT0, [PTR, #-13]
+ ldrbpl DAT0, [PTR, #-11]
+ addpl RESULT, RESULT, #2
+ teq DAT0, #0
+ it eq
+ subeq RESULT, RESULT, #1
+ b 98f
+96: add RESULT, RESULT, #4
+ teq TMP2, #0
+ beq 97f @ not in next 4 bytes
+ sub TMP2, DAT2, #0x20000
+ bics TMP2, TMP2, DAT2
+ itee mi
+ ldrbmi DAT0, [PTR, #-9]
+ ldrbpl DAT0, [PTR, #-7]
+ addpl RESULT, RESULT, #2
+ teq DAT0, #0
+ it eq
+ subeq RESULT, RESULT, #1
+ b 98f
+97: add RESULT, RESULT, #4
+ sub TMP3, DAT3, #0x20000
+ bics TMP3, TMP3, DAT3
+ itee mi
+ ldrbmi DAT0, [PTR, #-5]
+ ldrbpl DAT0, [PTR, #-3]
+ addpl RESULT, RESULT, #2
+ teq DAT0, #0
+ it eq
+ subeq RESULT, RESULT, #1
+ @ drop through to 98f
+98: setend le
+99: pop {v1-v6,pc}
+.endfunc
+
+ .unreq RESULT
+ .unreq BUF
+ .unreq SIZE
+ .unreq PATTERN
+ .unreq PTR
+ .unreq DAT0
+ .unreq DAT1
+ .unreq DAT2
+ .unreq DAT3
+ .unreq TMP0
+ .unreq TMP1
+ .unreq TMP2
+ .unreq TMP3
diff --git a/libavcodec/arm/h264dsp_init_arm.c b/libavcodec/arm/h264dsp_init_arm.c
index 785b604..2804e56 100644
--- a/libavcodec/arm/h264dsp_init_arm.c
+++ b/libavcodec/arm/h264dsp_init_arm.c
@@ -24,6 +24,8 @@
#include "libavutil/arm/cpu.h"
#include "libavcodec/h264dsp.h"
+int ff_h264_find_start_code_candidate_armv6(const uint8_t *buf, int size);
+
void ff_h264_v_loop_filter_luma_neon(uint8_t *pix, int stride, int alpha,
int beta, int8_t *tc0);
void ff_h264_h_loop_filter_luma_neon(uint8_t *pix, int stride, int alpha,
@@ -106,6 +108,8 @@ av_cold void ff_h264dsp_init_arm(H264DSPContext *c, const int bit_depth,
{
int cpu_flags = av_get_cpu_flags();
+ if (have_armv6(cpu_flags))
+ c->h264_find_start_code_candidate = ff_h264_find_start_code_candidate_armv6;
if (have_neon(cpu_flags))
ff_h264dsp_init_neon(c, bit_depth, chroma_format_idc);
}
diff --git a/libavcodec/h264_parser.c b/libavcodec/h264_parser.c
index 972aace..363843c 100644
--- a/libavcodec/h264_parser.c
+++ b/libavcodec/h264_parser.c
@@ -65,7 +65,6 @@ static int ff_h264_find_frame_end(H264Context *h, const uint8_t *buf, int buf_si
i += h->h264dsp.h264_find_start_code_candidate(buf + i, buf_size - i);
if (i < buf_size)
state = 2;
- }
}else if(state<=2){
if(buf[i]==1) state^= 5; //2->7, 1->4, 0->5
else if(buf[i]) state = 7;
--
1.8.5.1