[FFmpeg-devel] [PATCH] VP8 de/encode via libvpx
James Zern
jzern
Thu May 20 01:57:13 CEST 2010
On Wed, May 19, 2010 at 18:22, Diego Biurrun <diego at biurrun.de> wrote:
> On Wed, May 19, 2010 at 02:28:10PM -0400, James Zern wrote:
>>
>> The attached will enable decode and basic encode of VP8 through
>> libvpx.
>
> Some general comments about your patches:
>
As suggested I'll focus on the decoder first. Updated patch attached
based on all the comments, I no doubt missed a few things though.
> - use text/plain or text/x-diff as content type, if gmail sucks too
> ?much for this, get a decent mailer (or get gmail fixed)
> - leave empty lines between your reply and quoted text
> - use K&R style, 4 spaces indentation, no tabs, no trailing whitespace
> ?in the new files you add
> - split your files logically, not file by file, the latter is a big
> ?annoyance
> - try to get the decoder merged first, then submit the encoder
>
Done. indent -kr applied and others should be addressed.
>> Some notes and questions regarding the sources:
>> ?- libvpx{enc,dec} currently have a BSD style license, if it doesn't
>> make sense in this context we can go to LGPL
>
> Given that you have a longish license + patent clause and given
> that it makes no difference in practice I suggest you use
> FFmpeg's standard LGPL header.
>
Done.
>
>> --- /dev/null 2010-02-26 16:50:52.000000000 -0500
>> +++ libavcodec/libvpxdec.c ? ?2010-05-17 23:46:16.000000000 -0400
>> @@ -0,0 +1,168 @@
>> +
>> +/*!\file
>> + ? \brief VP8 decoder support via libvpx
>> +*/
>
> The Javadoc-style Doxygen variant is preferred.
>
Done.
>> +#include "avcodec.h"
>
> system headers before local headers please
>
Done.
>> +#ifndef HAVE_STDINT_H
>> +# define HAVE_STDINT_H 1
>> +#endif
>
> WTH?
>
Removed, was a holdover for mplayer building. The library currently
will define stdint types (much like EMULATE_INTTYPES used to) for
other platforms.
>> +#define VPX_CODEC_DISABLE_COMPAT 1
>> +#include <vpx_decoder.h>
>> +#include <vp8dx.h>
>> +
>> +#include <assert.h>
>> +
>> +typedef struct VP8DecoderContext {
>> + ?vpx_codec_ctx_t decoder;
>> +} vp8dctx_t;
>
> The _t namespace is reserved by POSIX, don't invade it.
>
Done.
>> +static av_cold int vp8_init(AVCodecContext* avctx)
>> +{
>> + ?vp8dctx_t* const ctx = avctx->priv_data;
>> + ?vpx_codec_iface_t* const iface = &vpx_codec_vp8_dx_algo;
>> + ?vpx_codec_dec_cfg_t deccfg = { /*token partitions+1 would be a decent choice*/
>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? .threads= FFMIN(avctx->thread_count,16) };
>> + ?vp8_postproc_cfg_t ppcfg;
>> +#if (defined(CONFIG_LIBPOSTPROC) && CONFIG_LIBPOSTPROC) ||\
>> + ? ?(defined(CONFIG_POSTPROC) && CONFIG_POSTPROC)
>
> I dunno who came up with this redundancy, but it's nonsense.
>
Removed post-processing for this go around.
>> + ? ?picture->data[0] = img->planes[0];
>> + ? ?picture->data[1] = img->planes[1];
>> + ? ?picture->data[2] = img->planes[2];
>> + ? ?picture->data[3] = img->planes[3];
>> + ? ?picture->linesize[0] = img->stride[0];
>> + ? ?picture->linesize[1] = img->stride[1];
>> + ? ?picture->linesize[2] = img->stride[2];
>> + ? ?picture->linesize[3] = img->stride[3];
>> + ? ?*data_size = sizeof(AVPicture);
>
> The = could be vertically aligned.
>
Done.
>> --- configure (revision 23157)
>> +++ configure (working copy)
>> @@ -183,6 +183,7 @@ External library support:
>> + ?--enable-libvpx-vp8 ? ? ?enable VP8 support via libvpx [no]
>
> Does the name of the lib really contain "vp8"? ?If not, then drop the
> silly "-vp8" postfix.
>
Removed.
>
>> --- doc/general.texi ?(revision 23153)
>> +++ doc/general.texi ?(working copy)
>> @@ -481,6 +481,8 @@ following image formats are supported:
>> ?@item V210 Quicktime Uncompressed 4:2:2 10-bit ? ? @tab ?X ?@tab ?X
>> ?@item VMware Screen Codec / VMware Video ?@tab ? ? @tab ?X
>> ? ? ?@tab Codec used in videos captured by VMware.
>> + at item VP8 ? ? ? ? ? ? ? ? ? ?@tab ?E ?@tab ?E
>> + ? ?@tab fourcc: VP80, encoding/decoding supported through external library libvpx
>
> Place it together with the other On2 codecs.
>
Done, but the name remains the same as that's how it's currently
publicly referenced.
-------------- next part --------------
Index: configure
===================================================================
--- configure (revision 23201)
+++ configure (working copy)
@@ -183,6 +183,7 @@
--enable-libtheora enable Theora encoding via libtheora [no]
--enable-libvorbis enable Vorbis encoding via libvorbis,
native implementation exists [no]
+ --enable-libvpx enable VP8 support via libvpx [no]
--enable-libx264 enable H.264 encoding via x264 [no]
--enable-libxvid enable Xvid encoding via xvidcore,
native MPEG-4/Xvid encoder exists [no]
@@ -927,6 +928,7 @@
libspeex
libtheora
libvorbis
+ libvpx
libx264
libxvid
lpc
@@ -1349,6 +1351,7 @@
libspeex_decoder_deps="libspeex"
libtheora_encoder_deps="libtheora"
libvorbis_encoder_deps="libvorbis"
+libvpx_decoder_deps="libvpx"
libx264_encoder_deps="libx264"
libxvid_encoder_deps="libxvid"
@@ -2617,6 +2620,7 @@
enabled libspeex && require libspeex speex/speex.h speex_decoder_init -lspeex
enabled libtheora && require libtheora theora/theoraenc.h th_info_init -ltheoraenc -ltheoradec -logg
enabled libvorbis && require libvorbis vorbis/vorbisenc.h vorbis_info_init -lvorbisenc -lvorbis -logg
+enabled libvpx && require2 libvpx "vpx_decoder.h vp8dx.h" "vpx_codec_dec_init_ver" -lvpx
enabled libx264 && require libx264 x264.h x264_encoder_encode -lx264 -lm &&
{ check_cpp_condition x264.h "X264_BUILD >= 90" ||
die "ERROR: libx264 version must be >= 0.90."; }
@@ -2886,6 +2890,7 @@
echo "libspeex enabled ${libspeex-no}"
echo "libtheora enabled ${libtheora-no}"
echo "libvorbis enabled ${libvorbis-no}"
+echo "libvpx enabled ${libvpx-no}"
echo "libx264 enabled ${libx264-no}"
echo "libxvid enabled ${libxvid-no}"
echo "zlib enabled ${zlib-no}"
Index: Changelog
===================================================================
--- Changelog (revision 23201)
+++ Changelog (working copy)
@@ -71,6 +71,7 @@
- spectral extension support in the E-AC-3 decoder
- unsharp video filter
- RTP hinting in the mov/3gp/mp4 muxer
+- VP8 decoding via libvpx
Index: libavcodec/libvpxdec.c
===================================================================
--- libavcodec/libvpxdec.c (revision 0)
+++ libavcodec/libvpxdec.c (revision 0)
@@ -0,0 +1,127 @@
+/*
+ * Copyright (c) 2010, Google, Inc.
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * @brief VP8 decoder support via libvpx
+ */
+
+#define VPX_CODEC_DISABLE_COMPAT 1
+#include <vpx_decoder.h>
+#include <vp8dx.h>
+
+#include "avcodec.h"
+
+typedef struct VP8DecoderContext {
+ vpx_codec_ctx_t decoder;
+} VP8Context;
+
+static av_cold int vp8_init(AVCodecContext *avctx)
+{
+ VP8Context *ctx = avctx->priv_data;
+ vpx_codec_iface_t *iface = &vpx_codec_vp8_dx_algo;
+ vpx_codec_dec_cfg_t deccfg = { /*token partitions+1 would be a decent choice */
+ .threads = FFMIN(avctx->thread_count, 16)
+ };
+
+ av_log(avctx, AV_LOG_INFO, "%s\n", vpx_codec_version_str());
+ av_log(avctx, AV_LOG_VERBOSE, "%s\n", vpx_codec_build_config());
+
+ if (vpx_codec_dec_init(&ctx->decoder, iface, &deccfg, 0) != VPX_CODEC_OK) {
+ const char *error = vpx_codec_error(&ctx->decoder);
+ av_log(avctx, AV_LOG_ERROR, "Failed to initialize decoder: %s\n",
+ error);
+ return -1;
+ }
+
+ avctx->pix_fmt = PIX_FMT_YUV420P;
+ return 0;
+}
+
+static int vp8_decode(AVCodecContext *avctx,
+ void *data, int *data_size, AVPacket *avpkt)
+{
+ VP8Context *ctx = avctx->priv_data;
+ AVFrame *picture = data;
+ vpx_codec_iter_t iter = NULL;
+ vpx_image_t *img;
+
+ /*end of stream, vp8 does no buffering */
+ if (avpkt->size == 0)
+ return 0;
+
+ if (vpx_codec_decode(&ctx->decoder, avpkt->data, avpkt->size, NULL, 0) !=
+ VPX_CODEC_OK) {
+ const char *error = vpx_codec_error(&ctx->decoder);
+ const char *detail = vpx_codec_error_detail(&ctx->decoder);
+
+ av_log(avctx, AV_LOG_ERROR, "Failed to decode frame: %s\n", error);
+ if (detail)
+ av_log(avctx, AV_LOG_ERROR, " Additional information: %s\n",
+ detail);
+ return -1;
+ }
+
+ if ((img = vpx_codec_get_frame(&ctx->decoder, &iter))) {
+ if (img->fmt != IMG_FMT_I420) {
+ av_log(avctx, AV_LOG_ERROR, "Unsupported output colorspace (%d)\n",
+ img->fmt);
+ return -1;
+ }
+
+ if ((int) img->d_w != avctx->width || (int) img->d_h != avctx->height) {
+ av_log(avctx, AV_LOG_INFO, "dimension change! %dx%d -> %dx%d\n",
+ avctx->width, avctx->height, img->d_w, img->d_h);
+ if (avcodec_check_dimensions(avctx, img->d_w, img->d_h))
+ return -1;
+ avcodec_set_dimensions(avctx, img->d_w, img->d_h);
+ }
+ picture->data[0] = img->planes[0];
+ picture->data[1] = img->planes[1];
+ picture->data[2] = img->planes[2];
+ picture->data[3] = NULL;
+ picture->linesize[0] = img->stride[0];
+ picture->linesize[1] = img->stride[1];
+ picture->linesize[2] = img->stride[2];
+ picture->linesize[3] = 0;
+ *data_size = sizeof(AVPicture);
+ }
+ return avpkt->size;
+}
+
+static av_cold int vp8_free(AVCodecContext *avctx)
+{
+ VP8Context *ctx = avctx->priv_data;
+ vpx_codec_destroy(&ctx->decoder);
+ return 0;
+}
+
+AVCodec libvpx_vp8_decoder = {
+ "libvpx_vp8",
+ AVMEDIA_TYPE_VIDEO,
+ CODEC_ID_VP8,
+ sizeof(VP8Context),
+ vp8_init,
+ NULL, /*encode */
+ vp8_free,
+ vp8_decode,
+ 0, /*capabilities */
+ .long_name = NULL_IF_CONFIG_SMALL("libvpx VP8"),
+};
Index: libavcodec/allcodecs.c
===================================================================
--- libavcodec/allcodecs.c (revision 23201)
+++ libavcodec/allcodecs.c (working copy)
@@ -349,6 +349,7 @@
REGISTER_DECODER (LIBSPEEX, libspeex);
REGISTER_ENCODER (LIBTHEORA, libtheora);
REGISTER_ENCODER (LIBVORBIS, libvorbis);
+ REGISTER_DECODER (LIBVPX_VP8, libvpx_vp8);
REGISTER_ENCODER (LIBX264, libx264);
REGISTER_ENCODER (LIBXVID, libxvid);
Index: libavcodec/avcodec.h
===================================================================
--- libavcodec/avcodec.h (revision 23201)
+++ libavcodec/avcodec.h (working copy)
@@ -30,7 +30,7 @@
#include "libavutil/avutil.h"
#define LIBAVCODEC_VERSION_MAJOR 52
-#define LIBAVCODEC_VERSION_MINOR 68
+#define LIBAVCODEC_VERSION_MINOR 69
#define LIBAVCODEC_VERSION_MICRO 2
#define LIBAVCODEC_VERSION_INT AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \
Index: libavcodec/Makefile
===================================================================
--- libavcodec/Makefile (revision 23201)
+++ libavcodec/Makefile (working copy)
@@ -536,6 +536,7 @@
OBJS-$(CONFIG_LIBSPEEX_DECODER) += libspeexdec.o
OBJS-$(CONFIG_LIBTHEORA_ENCODER) += libtheoraenc.o
OBJS-$(CONFIG_LIBVORBIS_ENCODER) += libvorbis.o
+OBJS-$(CONFIG_LIBVPX_VP8_DECODER) += libvpxdec.o
OBJS-$(CONFIG_LIBX264_ENCODER) += libx264.o
OBJS-$(CONFIG_LIBXVID_ENCODER) += libxvidff.o libxvid_rc.o
Index: doc/general.texi
===================================================================
--- doc/general.texi (revision 23201)
+++ doc/general.texi (working copy)
@@ -436,6 +436,8 @@
@tab fourcc: VP50
@item On2 VP6 @tab @tab X
@tab fourcc: VP60,VP61,VP62
+ at item VP8 @tab @tab X
+ @tab fourcc: VP80, decoding supported through external library libvpx
@item planar RGB @tab @tab X
@tab fourcc: 8BPS
@item Q-team QPEG @tab @tab X
More information about the ffmpeg-devel
mailing list