1
0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-12-18 03:19:31 +02:00
FFmpeg/libavcodec/crystalhd.c

1226 lines
43 KiB
C
Raw Normal View History

/*
* - CrystalHD decoder module -
*
* Copyright(C) 2010,2011 Philip Langdale <ffmpeg.philipl@overt.org>
*
* 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
*/
/*
* - Principles of Operation -
*
* The CrystalHD decoder operates at the bitstream level - which is an even
* higher level than the decoding hardware you typically see in modern GPUs.
* This means it has a very simple interface, in principle. You feed demuxed
* packets in one end and get decoded picture (fields/frames) out the other.
*
* Of course, nothing is ever that simple. Due, at the very least, to b-frame
* dependencies in the supported formats, the hardware has a delay between
* when a packet goes in, and when a picture comes out. Furthermore, this delay
* is not just a function of time, but also one of the dependency on additional
* frames being fed into the decoder to satisfy the b-frame dependencies.
*
* As such, a pipeline will build up that is roughly equivalent to the required
* DPB for the file being played. If that was all it took, things would still
* be simple - so, of course, it isn't.
*
* The hardware has a way of indicating that a picture is ready to be copied out,
* but this is unreliable - and sometimes the attempt will still fail so, based
* on testing, the code will wait until 3 pictures are ready before starting
* to copy out - and this has the effect of extending the pipeline.
*
* Finally, while it is tempting to say that once the decoder starts outputting
* frames, the software should never fail to return a frame from a decode(),
* this is a hard assertion to make, because the stream may switch between
* differently encoded content (number of b-frames, interlacing, etc) which
* might require a longer pipeline than before. If that happened, you could
* deadlock trying to retrieve a frame that can't be decoded without feeding
* in additional packets.
*
* As such, the code will return in the event that a picture cannot be copied
* out, leading to an increase in the length of the pipeline. This in turn,
* means we have to be sensitive to the time it takes to decode a picture;
* We do not want to give up just because the hardware needed a little more
* time to prepare the picture! For this reason, there are delays included
* in the decode() path that ensure that, under normal conditions, the hardware
* will only fail to return a frame if it really needs additional packets to
* complete the decoding.
*
* Finally, to be explicit, we do not want the pipeline to grow without bound
* for two reasons: 1) The hardware can only buffer a finite number of packets,
* and 2) The client application may not be able to cope with arbitrarily long
* delays in the video path relative to the audio path. For example. MPlayer
* can only handle a 20 picture delay (although this is arbitrary, and needs
* to be extended to fully support the CrystalHD where the delay could be up
* to 32 pictures - consider PAFF H.264 content with 16 b-frames).
*/
/*****************************************************************************
* Includes
****************************************************************************/
#define _XOPEN_SOURCE 600
#include <inttypes.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <libcrystalhd/bc_dts_types.h>
#include <libcrystalhd/bc_dts_defs.h>
#include <libcrystalhd/libcrystalhd_if.h>
#include "avcodec.h"
#include "h264.h"
#include "libavutil/imgutils.h"
#include "libavutil/intreadwrite.h"
#include "libavutil/opt.h"
/** Timeout parameter passed to DtsProcOutput() in us */
#define OUTPUT_PROC_TIMEOUT 50
/** Step between fake timestamps passed to hardware in units of 100ns */
#define TIMESTAMP_UNIT 100000
/** Initial value in us of the wait in decode() */
#define BASE_WAIT 10000
/** Increment in us to adjust wait in decode() */
#define WAIT_UNIT 1000
/*****************************************************************************
* Module private data
****************************************************************************/
typedef enum {
RET_ERROR = -1,
RET_OK = 0,
RET_COPY_AGAIN = 1,
RET_SKIP_NEXT_COPY = 2,
RET_COPY_NEXT_FIELD = 3,
} CopyRet;
typedef struct OpaqueList {
struct OpaqueList *next;
uint64_t fake_timestamp;
uint64_t reordered_opaque;
uint8_t pic_type;
} OpaqueList;
typedef struct {
AVClass *av_class;
AVCodecContext *avctx;
AVFrame pic;
HANDLE dev;
uint8_t *orig_extradata;
uint32_t orig_extradata_size;
AVBitStreamFilterContext *bsfc;
AVCodecParserContext *parser;
uint8_t is_70012;
uint8_t *sps_pps_buf;
uint32_t sps_pps_size;
uint8_t is_nal;
uint8_t output_ready;
uint8_t need_second_field;
uint8_t skip_next_output;
uint64_t decode_wait;
uint64_t last_picture;
OpaqueList *head;
OpaqueList *tail;
/* Options */
uint32_t sWidth;
uint8_t bframe_bug;
} CHDContext;
static const AVOption options[] = {
{ "crystalhd_downscale_width",
"Turn on downscaling to the specified width",
offsetof(CHDContext, sWidth),
AV_OPT_TYPE_INT, 0, 0, UINT32_MAX,
AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_DECODING_PARAM, },
{ NULL, },
};
/*****************************************************************************
* Helper functions
****************************************************************************/
static inline BC_MEDIA_SUBTYPE id2subtype(CHDContext *priv, enum CodecID id)
{
switch (id) {
case CODEC_ID_MPEG4:
return BC_MSUBTYPE_DIVX;
case CODEC_ID_MSMPEG4V3:
return BC_MSUBTYPE_DIVX311;
case CODEC_ID_MPEG2VIDEO:
return BC_MSUBTYPE_MPEG2VIDEO;
case CODEC_ID_VC1:
return BC_MSUBTYPE_VC1;
case CODEC_ID_WMV3:
return BC_MSUBTYPE_WMV3;
case CODEC_ID_H264:
return priv->is_nal ? BC_MSUBTYPE_AVC1 : BC_MSUBTYPE_H264;
default:
return BC_MSUBTYPE_INVALID;
}
}
static inline void print_frame_info(CHDContext *priv, BC_DTS_PROC_OUT *output)
{
av_log(priv->avctx, AV_LOG_VERBOSE, "\tYBuffSz: %u\n", output->YbuffSz);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tYBuffDoneSz: %u\n",
output->YBuffDoneSz);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tUVBuffDoneSz: %u\n",
output->UVBuffDoneSz);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tTimestamp: %"PRIu64"\n",
output->PicInfo.timeStamp);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tPicture Number: %u\n",
output->PicInfo.picture_number);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tWidth: %u\n",
output->PicInfo.width);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tHeight: %u\n",
output->PicInfo.height);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tChroma: 0x%03x\n",
output->PicInfo.chroma_format);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tPulldown: %u\n",
output->PicInfo.pulldown);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tFlags: 0x%08x\n",
output->PicInfo.flags);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tFrame Rate/Res: %u\n",
output->PicInfo.frame_rate);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tAspect Ratio: %u\n",
output->PicInfo.aspect_ratio);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tColor Primaries: %u\n",
output->PicInfo.colour_primaries);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tMetaData: %u\n",
output->PicInfo.picture_meta_payload);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tSession Number: %u\n",
output->PicInfo.sess_num);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tycom: %u\n",
output->PicInfo.ycom);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tCustom Aspect: %u\n",
output->PicInfo.custom_aspect_ratio_width_height);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tFrames to Drop: %u\n",
output->PicInfo.n_drop);
av_log(priv->avctx, AV_LOG_VERBOSE, "\tH264 Valid Fields: 0x%08x\n",
output->PicInfo.other.h264.valid);
}
/*****************************************************************************
* OpaqueList functions
****************************************************************************/
static uint64_t opaque_list_push(CHDContext *priv, uint64_t reordered_opaque,
uint8_t pic_type)
{
OpaqueList *newNode = av_mallocz(sizeof (OpaqueList));
if (!newNode) {
av_log(priv->avctx, AV_LOG_ERROR,
"Unable to allocate new node in OpaqueList.\n");
return 0;
}
if (!priv->head) {
newNode->fake_timestamp = TIMESTAMP_UNIT;
priv->head = newNode;
} else {
newNode->fake_timestamp = priv->tail->fake_timestamp + TIMESTAMP_UNIT;
priv->tail->next = newNode;
}
priv->tail = newNode;
newNode->reordered_opaque = reordered_opaque;
newNode->pic_type = pic_type;
return newNode->fake_timestamp;
}
/*
* The OpaqueList is built in decode order, while elements will be removed
* in presentation order. If frames are reordered, this means we must be
* able to remove elements that are not the first element.
*
* Returned node must be freed by caller.
*/
static OpaqueList *opaque_list_pop(CHDContext *priv, uint64_t fake_timestamp)
{
OpaqueList *node = priv->head;
if (!priv->head) {
av_log(priv->avctx, AV_LOG_ERROR,
"CrystalHD: Attempted to query non-existent timestamps.\n");
return NULL;
}
/*
* The first element is special-cased because we have to manipulate
* the head pointer rather than the previous element in the list.
*/
if (priv->head->fake_timestamp == fake_timestamp) {
priv->head = node->next;
if (!priv->head->next)
priv->tail = priv->head;
node->next = NULL;
return node;
}
/*
* The list is processed at arm's length so that we have the
* previous element available to rewrite its next pointer.
*/
while (node->next) {
OpaqueList *current = node->next;
if (current->fake_timestamp == fake_timestamp) {
node->next = current->next;
if (!node->next)
priv->tail = node;
current->next = NULL;
return current;
} else {
node = current;
}
}
av_log(priv->avctx, AV_LOG_VERBOSE,
"CrystalHD: Couldn't match fake_timestamp.\n");
return NULL;
}
/*****************************************************************************
* Video decoder API function definitions
****************************************************************************/
static void flush(AVCodecContext *avctx)
{
CHDContext *priv = avctx->priv_data;
avctx->has_b_frames = 0;
priv->last_picture = -1;
priv->output_ready = 0;
priv->need_second_field = 0;
priv->skip_next_output = 0;
priv->decode_wait = BASE_WAIT;
if (priv->pic.data[0])
avctx->release_buffer(avctx, &priv->pic);
/* Flush mode 4 flushes all software and hardware buffers. */
DtsFlushInput(priv->dev, 4);
}
static av_cold int uninit(AVCodecContext *avctx)
{
CHDContext *priv = avctx->priv_data;
HANDLE device;
device = priv->dev;
DtsStopDecoder(device);
DtsCloseDecoder(device);
DtsDeviceClose(device);
/*
* Restore original extradata, so that if the decoder is
* reinitialised, the bitstream detection and filtering
* will work as expected.
*/
if (priv->orig_extradata) {
av_free(avctx->extradata);
avctx->extradata = priv->orig_extradata;
avctx->extradata_size = priv->orig_extradata_size;
priv->orig_extradata = NULL;
priv->orig_extradata_size = 0;
}
av_parser_close(priv->parser);
if (priv->bsfc) {
av_bitstream_filter_close(priv->bsfc);
}
av_free(priv->sps_pps_buf);
if (priv->pic.data[0])
avctx->release_buffer(avctx, &priv->pic);
if (priv->head) {
OpaqueList *node = priv->head;
while (node) {
OpaqueList *next = node->next;
av_free(node);
node = next;
}
}
return 0;
}
static av_cold int init(AVCodecContext *avctx)
{
CHDContext* priv;
BC_STATUS ret;
BC_INFO_CRYSTAL version;
BC_INPUT_FORMAT format = {
.FGTEnable = FALSE,
.Progressive = TRUE,
.OptFlags = 0x80000000 | vdecFrameRate59_94 | 0x40,
.width = avctx->width,
.height = avctx->height,
};
BC_MEDIA_SUBTYPE subtype;
uint32_t mode = DTS_PLAYBACK_MODE |
DTS_LOAD_FILE_PLAY_FW |
DTS_SKIP_TX_CHK_CPB |
DTS_PLAYBACK_DROP_RPT_MODE |
DTS_SINGLE_THREADED_MODE |
DTS_DFLT_RESOLUTION(vdecRESOLUTION_1080p23_976);
av_log(avctx, AV_LOG_VERBOSE, "CrystalHD Init for %s\n",
avctx->codec->name);
avctx->pix_fmt = PIX_FMT_YUYV422;
/* Initialize the library */
priv = avctx->priv_data;
priv->avctx = avctx;
priv->is_nal = avctx->extradata_size > 0 && *(avctx->extradata) == 1;
priv->last_picture = -1;
priv->decode_wait = BASE_WAIT;
subtype = id2subtype(priv, avctx->codec->id);
switch (subtype) {
case BC_MSUBTYPE_AVC1:
{
uint8_t *dummy_p;
int dummy_int;
/* Back up the extradata so it can be restored at close time. */
priv->orig_extradata = av_malloc(avctx->extradata_size);
if (!priv->orig_extradata) {
av_log(avctx, AV_LOG_ERROR,
"Failed to allocate copy of extradata\n");
return AVERROR(ENOMEM);
}
priv->orig_extradata_size = avctx->extradata_size;
memcpy(priv->orig_extradata, avctx->extradata, avctx->extradata_size);
priv->bsfc = av_bitstream_filter_init("h264_mp4toannexb");
if (!priv->bsfc) {
av_log(avctx, AV_LOG_ERROR,
"Cannot open the h264_mp4toannexb BSF!\n");
return AVERROR_BSF_NOT_FOUND;
}
av_bitstream_filter_filter(priv->bsfc, avctx, NULL, &dummy_p,
&dummy_int, NULL, 0, 0);
}
subtype = BC_MSUBTYPE_H264;
// Fall-through
case BC_MSUBTYPE_H264:
format.startCodeSz = 4;
// Fall-through
case BC_MSUBTYPE_VC1:
case BC_MSUBTYPE_WVC1:
case BC_MSUBTYPE_WMV3:
case BC_MSUBTYPE_WMVA:
case BC_MSUBTYPE_MPEG2VIDEO:
case BC_MSUBTYPE_DIVX:
case BC_MSUBTYPE_DIVX311:
format.pMetaData = avctx->extradata;
format.metaDataSz = avctx->extradata_size;
break;
default:
av_log(avctx, AV_LOG_ERROR, "CrystalHD: Unknown codec name\n");
return AVERROR(EINVAL);
}
format.mSubtype = subtype;
if (priv->sWidth) {
format.bEnableScaling = 1;
format.ScalingParams.sWidth = priv->sWidth;
}
/* Get a decoder instance */
av_log(avctx, AV_LOG_VERBOSE, "CrystalHD: starting up\n");
// Initialize the Link and Decoder devices
ret = DtsDeviceOpen(&priv->dev, mode);
if (ret != BC_STS_SUCCESS) {
av_log(avctx, AV_LOG_VERBOSE, "CrystalHD: DtsDeviceOpen failed\n");
goto fail;
}
ret = DtsCrystalHDVersion(priv->dev, &version);
if (ret != BC_STS_SUCCESS) {
av_log(avctx, AV_LOG_VERBOSE,
"CrystalHD: DtsCrystalHDVersion failed\n");
goto fail;
}
priv->is_70012 = version.device == 0;
if (priv->is_70012 &&
(subtype == BC_MSUBTYPE_DIVX || subtype == BC_MSUBTYPE_DIVX311)) {
av_log(avctx, AV_LOG_VERBOSE,
"CrystalHD: BCM70012 doesn't support MPEG4-ASP/DivX/Xvid\n");
goto fail;
}
ret = DtsSetInputFormat(priv->dev, &format);
if (ret != BC_STS_SUCCESS) {
av_log(avctx, AV_LOG_ERROR, "CrystalHD: SetInputFormat failed\n");
goto fail;
}
ret = DtsOpenDecoder(priv->dev, BC_STREAM_TYPE_ES);
if (ret != BC_STS_SUCCESS) {
av_log(avctx, AV_LOG_ERROR, "CrystalHD: DtsOpenDecoder failed\n");
goto fail;
}
ret = DtsSetColorSpace(priv->dev, OUTPUT_MODE422_YUY2);
if (ret != BC_STS_SUCCESS) {
av_log(avctx, AV_LOG_ERROR, "CrystalHD: DtsSetColorSpace failed\n");
goto fail;
}
ret = DtsStartDecoder(priv->dev);
if (ret != BC_STS_SUCCESS) {
av_log(avctx, AV_LOG_ERROR, "CrystalHD: DtsStartDecoder failed\n");
goto fail;
}
ret = DtsStartCapture(priv->dev);
if (ret != BC_STS_SUCCESS) {
av_log(avctx, AV_LOG_ERROR, "CrystalHD: DtsStartCapture failed\n");
goto fail;
}
if (avctx->codec->id == CODEC_ID_H264) {
priv->parser = av_parser_init(avctx->codec->id);
if (!priv->parser)
av_log(avctx, AV_LOG_WARNING,
"Cannot open the h.264 parser! Interlaced h.264 content "
"will not be detected reliably.\n");
priv->parser->flags = PARSER_FLAG_COMPLETE_FRAMES;
}
av_log(avctx, AV_LOG_VERBOSE, "CrystalHD: Init complete.\n");
return 0;
fail:
uninit(avctx);
return -1;
}
static inline CopyRet copy_frame(AVCodecContext *avctx,
BC_DTS_PROC_OUT *output,
CrystalHD: Improve detection of h.264 content. As previously discussed, the CrystalHD hardware returns exceptionally useless information about interlaced h.264 content - to the extent that it's not possible to distinguish most MBAFF and PAFF content until it's too late. In an attempt to compensate for this, I'm introducing two mechanisms: 1) Peeking at the picture number of the next picture The hardware provides a capability to peek the next picture number. If it is the same as the current picture number, then we are clearly dealing with two fields and not a frame or fieldpair. If this always worked, it would be all we need, but it's not guaranteed to work. Sometimes, the next picture may not be decoded sufficiently for the number to be known; alternately, a corruption in the stream may cause the hardware to refuse to return the number even if the next intact frame is decoded. In either case, the query will return 0. If we are unable to peek the next picture number, we assume that the picture is a frame/fieldpair and return it accordingly. If that turns out to be incorrect, we discard the second field, and the user has to live with the glitch. In testing, false detection can occur for the first couple of seconds, and then the pipeline stabalizes and we get correct detection. 2) Use the h264_parser to detect when individual input fields have been combined into an output fieldpair. I have multiple PAFF samples where this behaviour is detected. The peeking mechanism described above will correctly detect that the output is a fieldpair, but we need to know what the input type was to ensure pipeline stability (only return one output frame per input frame). If we find ourselves with an output fieldpair, yet the input picture type was a field, as reported by the parser, then we are dealing with this case, and can make sure not to return anything on the next decode() call. Taken together, these allow us to remove the hard-coded hacks for different h.264 types, and we can clearly describe the conditions under which we can trust the hardware's claim that content is interlaced. Signed-off-by: Philip Langdale <philipl@overt.org>
2011-03-26 19:34:20 +02:00
void *data, int *data_size)
{
BC_STATUS ret;
BC_DTS_STATUS decoder_status = { 0, };
CrystalHD: Improve detection of h.264 content. As previously discussed, the CrystalHD hardware returns exceptionally useless information about interlaced h.264 content - to the extent that it's not possible to distinguish most MBAFF and PAFF content until it's too late. In an attempt to compensate for this, I'm introducing two mechanisms: 1) Peeking at the picture number of the next picture The hardware provides a capability to peek the next picture number. If it is the same as the current picture number, then we are clearly dealing with two fields and not a frame or fieldpair. If this always worked, it would be all we need, but it's not guaranteed to work. Sometimes, the next picture may not be decoded sufficiently for the number to be known; alternately, a corruption in the stream may cause the hardware to refuse to return the number even if the next intact frame is decoded. In either case, the query will return 0. If we are unable to peek the next picture number, we assume that the picture is a frame/fieldpair and return it accordingly. If that turns out to be incorrect, we discard the second field, and the user has to live with the glitch. In testing, false detection can occur for the first couple of seconds, and then the pipeline stabalizes and we get correct detection. 2) Use the h264_parser to detect when individual input fields have been combined into an output fieldpair. I have multiple PAFF samples where this behaviour is detected. The peeking mechanism described above will correctly detect that the output is a fieldpair, but we need to know what the input type was to ensure pipeline stability (only return one output frame per input frame). If we find ourselves with an output fieldpair, yet the input picture type was a field, as reported by the parser, then we are dealing with this case, and can make sure not to return anything on the next decode() call. Taken together, these allow us to remove the hard-coded hacks for different h.264 types, and we can clearly describe the conditions under which we can trust the hardware's claim that content is interlaced. Signed-off-by: Philip Langdale <philipl@overt.org>
2011-03-26 19:34:20 +02:00
uint8_t trust_interlaced;
uint8_t interlaced;
CHDContext *priv = avctx->priv_data;
int64_t pkt_pts = AV_NOPTS_VALUE;
uint8_t pic_type = 0;
uint8_t bottom_field = (output->PicInfo.flags & VDEC_FLAG_BOTTOMFIELD) ==
VDEC_FLAG_BOTTOMFIELD;
uint8_t bottom_first = !!(output->PicInfo.flags & VDEC_FLAG_BOTTOM_FIRST);
int width = output->PicInfo.width;
int height = output->PicInfo.height;
int bwidth;
uint8_t *src = output->Ybuff;
int sStride;
uint8_t *dst;
int dStride;
if (output->PicInfo.timeStamp != 0) {
OpaqueList *node = opaque_list_pop(priv, output->PicInfo.timeStamp);
if (node) {
pkt_pts = node->reordered_opaque;
pic_type = node->pic_type;
av_free(node);
} else {
/*
* We will encounter a situation where a timestamp cannot be
* popped if a second field is being returned. In this case,
* each field has the same timestamp and the first one will
* cause it to be popped. To keep subsequent calculations
* simple, pic_type should be set a FIELD value - doesn't
* matter which, but I chose BOTTOM.
*/
pic_type = PICT_BOTTOM_FIELD;
}
av_log(avctx, AV_LOG_VERBOSE, "output \"pts\": %"PRIu64"\n",
output->PicInfo.timeStamp);
av_log(avctx, AV_LOG_VERBOSE, "output picture type %d\n",
pic_type);
}
ret = DtsGetDriverStatus(priv->dev, &decoder_status);
if (ret != BC_STS_SUCCESS) {
av_log(avctx, AV_LOG_ERROR,
"CrystalHD: GetDriverStatus failed: %u\n", ret);
return RET_ERROR;
}
CrystalHD: Improve detection of h.264 content. As previously discussed, the CrystalHD hardware returns exceptionally useless information about interlaced h.264 content - to the extent that it's not possible to distinguish most MBAFF and PAFF content until it's too late. In an attempt to compensate for this, I'm introducing two mechanisms: 1) Peeking at the picture number of the next picture The hardware provides a capability to peek the next picture number. If it is the same as the current picture number, then we are clearly dealing with two fields and not a frame or fieldpair. If this always worked, it would be all we need, but it's not guaranteed to work. Sometimes, the next picture may not be decoded sufficiently for the number to be known; alternately, a corruption in the stream may cause the hardware to refuse to return the number even if the next intact frame is decoded. In either case, the query will return 0. If we are unable to peek the next picture number, we assume that the picture is a frame/fieldpair and return it accordingly. If that turns out to be incorrect, we discard the second field, and the user has to live with the glitch. In testing, false detection can occur for the first couple of seconds, and then the pipeline stabalizes and we get correct detection. 2) Use the h264_parser to detect when individual input fields have been combined into an output fieldpair. I have multiple PAFF samples where this behaviour is detected. The peeking mechanism described above will correctly detect that the output is a fieldpair, but we need to know what the input type was to ensure pipeline stability (only return one output frame per input frame). If we find ourselves with an output fieldpair, yet the input picture type was a field, as reported by the parser, then we are dealing with this case, and can make sure not to return anything on the next decode() call. Taken together, these allow us to remove the hard-coded hacks for different h.264 types, and we can clearly describe the conditions under which we can trust the hardware's claim that content is interlaced. Signed-off-by: Philip Langdale <philipl@overt.org>
2011-03-26 19:34:20 +02:00
/*
* For most content, we can trust the interlaced flag returned
* by the hardware, but sometimes we can't. These are the
* conditions under which we can trust the flag:
*
* 1) It's not h.264 content
* 2) The UNKNOWN_SRC flag is not set
* 3) We know we're expecting a second field
* 4) The hardware reports this picture and the next picture
* have the same picture number.
*
* Note that there can still be interlaced content that will
* fail this check, if the hardware hasn't decoded the next
* picture or if there is a corruption in the stream. (In either
* case a 0 will be returned for the next picture number)
*/
trust_interlaced = avctx->codec->id != CODEC_ID_H264 ||
!(output->PicInfo.flags & VDEC_FLAG_UNKNOWN_SRC) ||
priv->need_second_field ||
(decoder_status.picNumFlags & ~0x40000000) ==
output->PicInfo.picture_number;
/*
* If we got a false negative for trust_interlaced on the first field,
* we will realise our mistake here when we see that the picture number is that
* of the previous picture. We cannot recover the frame and should discard the
* second field to keep the correct number of output frames.
*/
if (output->PicInfo.picture_number == priv->last_picture && !priv->need_second_field) {
av_log(avctx, AV_LOG_WARNING,
"Incorrectly guessed progressive frame. Discarding second field\n");
/* Returning without providing a picture. */
return RET_OK;
}
interlaced = (output->PicInfo.flags & VDEC_FLAG_INTERLACED_SRC) &&
trust_interlaced;
CrystalHD: Improve detection of h.264 content. As previously discussed, the CrystalHD hardware returns exceptionally useless information about interlaced h.264 content - to the extent that it's not possible to distinguish most MBAFF and PAFF content until it's too late. In an attempt to compensate for this, I'm introducing two mechanisms: 1) Peeking at the picture number of the next picture The hardware provides a capability to peek the next picture number. If it is the same as the current picture number, then we are clearly dealing with two fields and not a frame or fieldpair. If this always worked, it would be all we need, but it's not guaranteed to work. Sometimes, the next picture may not be decoded sufficiently for the number to be known; alternately, a corruption in the stream may cause the hardware to refuse to return the number even if the next intact frame is decoded. In either case, the query will return 0. If we are unable to peek the next picture number, we assume that the picture is a frame/fieldpair and return it accordingly. If that turns out to be incorrect, we discard the second field, and the user has to live with the glitch. In testing, false detection can occur for the first couple of seconds, and then the pipeline stabalizes and we get correct detection. 2) Use the h264_parser to detect when individual input fields have been combined into an output fieldpair. I have multiple PAFF samples where this behaviour is detected. The peeking mechanism described above will correctly detect that the output is a fieldpair, but we need to know what the input type was to ensure pipeline stability (only return one output frame per input frame). If we find ourselves with an output fieldpair, yet the input picture type was a field, as reported by the parser, then we are dealing with this case, and can make sure not to return anything on the next decode() call. Taken together, these allow us to remove the hard-coded hacks for different h.264 types, and we can clearly describe the conditions under which we can trust the hardware's claim that content is interlaced. Signed-off-by: Philip Langdale <philipl@overt.org>
2011-03-26 19:34:20 +02:00
if (!trust_interlaced && (decoder_status.picNumFlags & ~0x40000000) == 0) {
av_log(avctx, AV_LOG_VERBOSE,
"Next picture number unknown. Assuming progressive frame.\n");
}
av_log(avctx, AV_LOG_VERBOSE, "Interlaced state: %d | trust_interlaced %d\n",
interlaced, trust_interlaced);
if (priv->pic.data[0] && !priv->need_second_field)
avctx->release_buffer(avctx, &priv->pic);
priv->need_second_field = interlaced && !priv->need_second_field;
priv->pic.buffer_hints = FF_BUFFER_HINTS_VALID | FF_BUFFER_HINTS_PRESERVE |
FF_BUFFER_HINTS_REUSABLE;
if (!priv->pic.data[0]) {
if (avctx->get_buffer(avctx, &priv->pic) < 0) {
av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
return RET_ERROR;
}
}
bwidth = av_image_get_linesize(avctx->pix_fmt, width, 0);
if (priv->is_70012) {
int pStride;
if (width <= 720)
pStride = 720;
else if (width <= 1280)
pStride = 1280;
else pStride = 1920;
sStride = av_image_get_linesize(avctx->pix_fmt, pStride, 0);
} else {
sStride = bwidth;
}
dStride = priv->pic.linesize[0];
dst = priv->pic.data[0];
av_log(priv->avctx, AV_LOG_VERBOSE, "CrystalHD: Copying out frame\n");
if (interlaced) {
int dY = 0;
int sY = 0;
height /= 2;
if (bottom_field) {
av_log(priv->avctx, AV_LOG_VERBOSE, "Interlaced: bottom field\n");
dY = 1;
} else {
av_log(priv->avctx, AV_LOG_VERBOSE, "Interlaced: top field\n");
dY = 0;
}
for (sY = 0; sY < height; dY++, sY++) {
memcpy(&(dst[dY * dStride]), &(src[sY * sStride]), bwidth);
dY++;
}
} else {
av_image_copy_plane(dst, dStride, src, sStride, bwidth, height);
}
priv->pic.interlaced_frame = interlaced;
if (interlaced)
priv->pic.top_field_first = !bottom_first;
priv->pic.pkt_pts = pkt_pts;
if (!priv->need_second_field) {
*data_size = sizeof(AVFrame);
*(AVFrame *)data = priv->pic;
}
CrystalHD: Improve detection of h.264 content. As previously discussed, the CrystalHD hardware returns exceptionally useless information about interlaced h.264 content - to the extent that it's not possible to distinguish most MBAFF and PAFF content until it's too late. In an attempt to compensate for this, I'm introducing two mechanisms: 1) Peeking at the picture number of the next picture The hardware provides a capability to peek the next picture number. If it is the same as the current picture number, then we are clearly dealing with two fields and not a frame or fieldpair. If this always worked, it would be all we need, but it's not guaranteed to work. Sometimes, the next picture may not be decoded sufficiently for the number to be known; alternately, a corruption in the stream may cause the hardware to refuse to return the number even if the next intact frame is decoded. In either case, the query will return 0. If we are unable to peek the next picture number, we assume that the picture is a frame/fieldpair and return it accordingly. If that turns out to be incorrect, we discard the second field, and the user has to live with the glitch. In testing, false detection can occur for the first couple of seconds, and then the pipeline stabalizes and we get correct detection. 2) Use the h264_parser to detect when individual input fields have been combined into an output fieldpair. I have multiple PAFF samples where this behaviour is detected. The peeking mechanism described above will correctly detect that the output is a fieldpair, but we need to know what the input type was to ensure pipeline stability (only return one output frame per input frame). If we find ourselves with an output fieldpair, yet the input picture type was a field, as reported by the parser, then we are dealing with this case, and can make sure not to return anything on the next decode() call. Taken together, these allow us to remove the hard-coded hacks for different h.264 types, and we can clearly describe the conditions under which we can trust the hardware's claim that content is interlaced. Signed-off-by: Philip Langdale <philipl@overt.org>
2011-03-26 19:34:20 +02:00
/*
* Two types of PAFF content have been observed. One form causes the
* hardware to return a field pair and the other individual fields,
* even though the input is always individual fields. We must skip
* copying on the next decode() call to maintain pipeline length in
* the first case.
*/
if (!interlaced && (output->PicInfo.flags & VDEC_FLAG_UNKNOWN_SRC) &&
(pic_type == PICT_TOP_FIELD || pic_type == PICT_BOTTOM_FIELD)) {
av_log(priv->avctx, AV_LOG_VERBOSE, "Fieldpair from two packets.\n");
return RET_SKIP_NEXT_COPY;
}
/*
* The logic here is purely based on empirical testing with samples.
* If we need a second field, it could come from a second input packet,
* or it could come from the same field-pair input packet at the current
* field. In the first case, we should return and wait for the next time
* round to get the second field, while in the second case, we should
* ask the decoder for it immediately.
*
* Testing has shown that we are dealing with the fieldpair -> two fields
* case if the VDEC_FLAG_UNKNOWN_SRC is not set or if the input picture
* type was PICT_FRAME (in this second case, the flag might still be set)
*/
return priv->need_second_field &&
(!(output->PicInfo.flags & VDEC_FLAG_UNKNOWN_SRC) ||
pic_type == PICT_FRAME) ?
RET_COPY_NEXT_FIELD : RET_OK;
}
static inline CopyRet receive_frame(AVCodecContext *avctx,
CrystalHD: Improve detection of h.264 content. As previously discussed, the CrystalHD hardware returns exceptionally useless information about interlaced h.264 content - to the extent that it's not possible to distinguish most MBAFF and PAFF content until it's too late. In an attempt to compensate for this, I'm introducing two mechanisms: 1) Peeking at the picture number of the next picture The hardware provides a capability to peek the next picture number. If it is the same as the current picture number, then we are clearly dealing with two fields and not a frame or fieldpair. If this always worked, it would be all we need, but it's not guaranteed to work. Sometimes, the next picture may not be decoded sufficiently for the number to be known; alternately, a corruption in the stream may cause the hardware to refuse to return the number even if the next intact frame is decoded. In either case, the query will return 0. If we are unable to peek the next picture number, we assume that the picture is a frame/fieldpair and return it accordingly. If that turns out to be incorrect, we discard the second field, and the user has to live with the glitch. In testing, false detection can occur for the first couple of seconds, and then the pipeline stabalizes and we get correct detection. 2) Use the h264_parser to detect when individual input fields have been combined into an output fieldpair. I have multiple PAFF samples where this behaviour is detected. The peeking mechanism described above will correctly detect that the output is a fieldpair, but we need to know what the input type was to ensure pipeline stability (only return one output frame per input frame). If we find ourselves with an output fieldpair, yet the input picture type was a field, as reported by the parser, then we are dealing with this case, and can make sure not to return anything on the next decode() call. Taken together, these allow us to remove the hard-coded hacks for different h.264 types, and we can clearly describe the conditions under which we can trust the hardware's claim that content is interlaced. Signed-off-by: Philip Langdale <philipl@overt.org>
2011-03-26 19:34:20 +02:00
void *data, int *data_size)
{
BC_STATUS ret;
BC_DTS_PROC_OUT output = {
.PicInfo.width = avctx->width,
.PicInfo.height = avctx->height,
};
CHDContext *priv = avctx->priv_data;
HANDLE dev = priv->dev;
*data_size = 0;
// Request decoded data from the driver
ret = DtsProcOutputNoCopy(dev, OUTPUT_PROC_TIMEOUT, &output);
if (ret == BC_STS_FMT_CHANGE) {
av_log(avctx, AV_LOG_VERBOSE, "CrystalHD: Initial format change\n");
avctx->width = output.PicInfo.width;
avctx->height = output.PicInfo.height;
switch ( output.PicInfo.aspect_ratio ) {
case vdecAspectRatioSquare:
avctx->sample_aspect_ratio = (AVRational) { 1, 1};
break;
case vdecAspectRatio12_11:
avctx->sample_aspect_ratio = (AVRational) { 12, 11};
break;
case vdecAspectRatio10_11:
avctx->sample_aspect_ratio = (AVRational) { 10, 11};
break;
case vdecAspectRatio16_11:
avctx->sample_aspect_ratio = (AVRational) { 16, 11};
break;
case vdecAspectRatio40_33:
avctx->sample_aspect_ratio = (AVRational) { 40, 33};
break;
case vdecAspectRatio24_11:
avctx->sample_aspect_ratio = (AVRational) { 24, 11};
break;
case vdecAspectRatio20_11:
avctx->sample_aspect_ratio = (AVRational) { 20, 11};
break;
case vdecAspectRatio32_11:
avctx->sample_aspect_ratio = (AVRational) { 32, 11};
break;
case vdecAspectRatio80_33:
avctx->sample_aspect_ratio = (AVRational) { 80, 33};
break;
case vdecAspectRatio18_11:
avctx->sample_aspect_ratio = (AVRational) { 18, 11};
break;
case vdecAspectRatio15_11:
avctx->sample_aspect_ratio = (AVRational) { 15, 11};
break;
case vdecAspectRatio64_33:
avctx->sample_aspect_ratio = (AVRational) { 64, 33};
break;
case vdecAspectRatio160_99:
avctx->sample_aspect_ratio = (AVRational) {160, 99};
break;
case vdecAspectRatio4_3:
avctx->sample_aspect_ratio = (AVRational) { 4, 3};
break;
case vdecAspectRatio16_9:
avctx->sample_aspect_ratio = (AVRational) { 16, 9};
break;
case vdecAspectRatio221_1:
avctx->sample_aspect_ratio = (AVRational) {221, 1};
break;
}
return RET_COPY_AGAIN;
} else if (ret == BC_STS_SUCCESS) {
int copy_ret = -1;
if (output.PoutFlags & BC_POUT_FLAGS_PIB_VALID) {
if (priv->last_picture == -1) {
/*
* Init to one less, so that the incrementing code doesn't
* need to be special-cased.
*/
priv->last_picture = output.PicInfo.picture_number - 1;
}
if (avctx->codec->id == CODEC_ID_MPEG4 &&
output.PicInfo.timeStamp == 0 && priv->bframe_bug) {
av_log(avctx, AV_LOG_VERBOSE,
"CrystalHD: Not returning packed frame twice.\n");
priv->last_picture++;
DtsReleaseOutputBuffs(dev, NULL, FALSE);
return RET_COPY_AGAIN;
}
print_frame_info(priv, &output);
if (priv->last_picture + 1 < output.PicInfo.picture_number) {
av_log(avctx, AV_LOG_WARNING,
"CrystalHD: Picture Number discontinuity\n");
/*
* Have we lost frames? If so, we need to shrink the
* pipeline length appropriately.
*
* XXX: I have no idea what the semantics of this situation
* are so I don't even know if we've lost frames or which
* ones.
*
* In any case, only warn the first time.
*/
priv->last_picture = output.PicInfo.picture_number - 1;
}
CrystalHD: Improve detection of h.264 content. As previously discussed, the CrystalHD hardware returns exceptionally useless information about interlaced h.264 content - to the extent that it's not possible to distinguish most MBAFF and PAFF content until it's too late. In an attempt to compensate for this, I'm introducing two mechanisms: 1) Peeking at the picture number of the next picture The hardware provides a capability to peek the next picture number. If it is the same as the current picture number, then we are clearly dealing with two fields and not a frame or fieldpair. If this always worked, it would be all we need, but it's not guaranteed to work. Sometimes, the next picture may not be decoded sufficiently for the number to be known; alternately, a corruption in the stream may cause the hardware to refuse to return the number even if the next intact frame is decoded. In either case, the query will return 0. If we are unable to peek the next picture number, we assume that the picture is a frame/fieldpair and return it accordingly. If that turns out to be incorrect, we discard the second field, and the user has to live with the glitch. In testing, false detection can occur for the first couple of seconds, and then the pipeline stabalizes and we get correct detection. 2) Use the h264_parser to detect when individual input fields have been combined into an output fieldpair. I have multiple PAFF samples where this behaviour is detected. The peeking mechanism described above will correctly detect that the output is a fieldpair, but we need to know what the input type was to ensure pipeline stability (only return one output frame per input frame). If we find ourselves with an output fieldpair, yet the input picture type was a field, as reported by the parser, then we are dealing with this case, and can make sure not to return anything on the next decode() call. Taken together, these allow us to remove the hard-coded hacks for different h.264 types, and we can clearly describe the conditions under which we can trust the hardware's claim that content is interlaced. Signed-off-by: Philip Langdale <philipl@overt.org>
2011-03-26 19:34:20 +02:00
copy_ret = copy_frame(avctx, &output, data, data_size);
if (*data_size > 0) {
avctx->has_b_frames--;
priv->last_picture++;
av_log(avctx, AV_LOG_VERBOSE, "CrystalHD: Pipeline length: %u\n",
avctx->has_b_frames);
}
} else {
/*
* An invalid frame has been consumed.
*/
av_log(avctx, AV_LOG_ERROR, "CrystalHD: ProcOutput succeeded with "
"invalid PIB\n");
avctx->has_b_frames--;
copy_ret = RET_OK;
}
DtsReleaseOutputBuffs(dev, NULL, FALSE);
return copy_ret;
} else if (ret == BC_STS_BUSY) {
return RET_COPY_AGAIN;
} else {
av_log(avctx, AV_LOG_ERROR, "CrystalHD: ProcOutput failed %d\n", ret);
return RET_ERROR;
}
}
static int decode(AVCodecContext *avctx, void *data, int *data_size, AVPacket *avpkt)
{
BC_STATUS ret;
BC_DTS_STATUS decoder_status = { 0, };
CopyRet rec_ret;
CHDContext *priv = avctx->priv_data;
HANDLE dev = priv->dev;
uint8_t *in_data = avpkt->data;
int len = avpkt->size;
int free_data = 0;
uint8_t pic_type = 0;
av_log(avctx, AV_LOG_VERBOSE, "CrystalHD: decode_frame\n");
if (avpkt->size == 7 && !priv->bframe_bug) {
/*
* The use of a drop frame triggers the bug
*/
av_log(avctx, AV_LOG_INFO,
"CrystalHD: Enabling work-around for packed b-frame bug\n");
priv->bframe_bug = 1;
} else if (avpkt->size == 8 && priv->bframe_bug) {
/*
* Delay frames don't trigger the bug
*/
av_log(avctx, AV_LOG_INFO,
"CrystalHD: Disabling work-around for packed b-frame bug\n");
priv->bframe_bug = 0;
}
if (len) {
int32_t tx_free = (int32_t)DtsTxFreeSize(dev);
if (priv->parser) {
int ret = 0;
if (priv->bsfc) {
ret = av_bitstream_filter_filter(priv->bsfc, avctx, NULL,
&in_data, &len,
avpkt->data, len, 0);
}
free_data = ret > 0;
if (ret >= 0) {
uint8_t *pout;
int psize;
int index;
H264Context *h = priv->parser->priv_data;
index = av_parser_parse2(priv->parser, avctx, &pout, &psize,
in_data, len, avctx->pkt->pts,
avctx->pkt->dts, 0);
if (index < 0) {
av_log(avctx, AV_LOG_WARNING,
"CrystalHD: Failed to parse h.264 packet to "
"detect interlacing.\n");
} else if (index != len) {
av_log(avctx, AV_LOG_WARNING,
"CrystalHD: Failed to parse h.264 packet "
"completely. Interlaced frames may be "
"incorrectly detected\n.");
} else {
av_log(avctx, AV_LOG_VERBOSE,
"CrystalHD: parser picture type %d\n",
h->s.picture_structure);
pic_type = h->s.picture_structure;
}
} else {
av_log(avctx, AV_LOG_WARNING,
"CrystalHD: mp4toannexb filter failed to filter "
"packet. Interlaced frames may be incorrectly "
"detected.\n");
}
}
if (len < tx_free - 1024) {
/*
* Despite being notionally opaque, either libcrystalhd or
* the hardware itself will mangle pts values that are too
* small or too large. The docs claim it should be in units
* of 100ns. Given that we're nominally dealing with a black
* box on both sides, any transform we do has no guarantee of
* avoiding mangling so we need to build a mapping to values
* we know will not be mangled.
*/
uint64_t pts = opaque_list_push(priv, avctx->pkt->pts, pic_type);
if (!pts) {
if (free_data) {
av_freep(&in_data);
}
return AVERROR(ENOMEM);
}
av_log(priv->avctx, AV_LOG_VERBOSE,
"input \"pts\": %"PRIu64"\n", pts);
ret = DtsProcInput(dev, in_data, len, pts, 0);
if (free_data) {
av_freep(&in_data);
}
if (ret == BC_STS_BUSY) {
av_log(avctx, AV_LOG_WARNING,
"CrystalHD: ProcInput returned busy\n");
usleep(BASE_WAIT);
return AVERROR(EBUSY);
} else if (ret != BC_STS_SUCCESS) {
av_log(avctx, AV_LOG_ERROR,
"CrystalHD: ProcInput failed: %u\n", ret);
return -1;
}
avctx->has_b_frames++;
} else {
av_log(avctx, AV_LOG_WARNING, "CrystalHD: Input buffer full\n");
len = 0; // We didn't consume any bytes.
}
} else {
av_log(avctx, AV_LOG_INFO, "CrystalHD: No more input data\n");
}
if (priv->skip_next_output) {
av_log(avctx, AV_LOG_VERBOSE, "CrystalHD: Skipping next output.\n");
priv->skip_next_output = 0;
avctx->has_b_frames--;
return len;
}
ret = DtsGetDriverStatus(dev, &decoder_status);
if (ret != BC_STS_SUCCESS) {
av_log(avctx, AV_LOG_ERROR, "CrystalHD: GetDriverStatus failed\n");
return -1;
}
/*
* No frames ready. Don't try to extract.
*
* Empirical testing shows that ReadyListCount can be a damn lie,
* and ProcOut still fails when count > 0. The same testing showed
* that two more iterations were needed before ProcOutput would
* succeed.
*/
if (priv->output_ready < 2) {
if (decoder_status.ReadyListCount != 0)
priv->output_ready++;
usleep(BASE_WAIT);
av_log(avctx, AV_LOG_INFO, "CrystalHD: Filling pipeline.\n");
return len;
} else if (decoder_status.ReadyListCount == 0) {
/*
* After the pipeline is established, if we encounter a lack of frames
* that probably means we're not giving the hardware enough time to
* decode them, so start increasing the wait time at the end of a
* decode call.
*/
usleep(BASE_WAIT);
priv->decode_wait += WAIT_UNIT;
av_log(avctx, AV_LOG_INFO, "CrystalHD: No frames ready. Returning\n");
return len;
}
do {
CrystalHD: Improve detection of h.264 content. As previously discussed, the CrystalHD hardware returns exceptionally useless information about interlaced h.264 content - to the extent that it's not possible to distinguish most MBAFF and PAFF content until it's too late. In an attempt to compensate for this, I'm introducing two mechanisms: 1) Peeking at the picture number of the next picture The hardware provides a capability to peek the next picture number. If it is the same as the current picture number, then we are clearly dealing with two fields and not a frame or fieldpair. If this always worked, it would be all we need, but it's not guaranteed to work. Sometimes, the next picture may not be decoded sufficiently for the number to be known; alternately, a corruption in the stream may cause the hardware to refuse to return the number even if the next intact frame is decoded. In either case, the query will return 0. If we are unable to peek the next picture number, we assume that the picture is a frame/fieldpair and return it accordingly. If that turns out to be incorrect, we discard the second field, and the user has to live with the glitch. In testing, false detection can occur for the first couple of seconds, and then the pipeline stabalizes and we get correct detection. 2) Use the h264_parser to detect when individual input fields have been combined into an output fieldpair. I have multiple PAFF samples where this behaviour is detected. The peeking mechanism described above will correctly detect that the output is a fieldpair, but we need to know what the input type was to ensure pipeline stability (only return one output frame per input frame). If we find ourselves with an output fieldpair, yet the input picture type was a field, as reported by the parser, then we are dealing with this case, and can make sure not to return anything on the next decode() call. Taken together, these allow us to remove the hard-coded hacks for different h.264 types, and we can clearly describe the conditions under which we can trust the hardware's claim that content is interlaced. Signed-off-by: Philip Langdale <philipl@overt.org>
2011-03-26 19:34:20 +02:00
rec_ret = receive_frame(avctx, data, data_size);
if (rec_ret == RET_OK && *data_size == 0) {
/*
* This case is for when the encoded fields are stored
* separately and we get a separate avpkt for each one. To keep
* the pipeline stable, we should return nothing and wait for
* the next time round to grab the second field.
* H.264 PAFF is an example of this.
*/
av_log(avctx, AV_LOG_VERBOSE, "Returning after first field.\n");
avctx->has_b_frames--;
} else if (rec_ret == RET_COPY_NEXT_FIELD) {
/*
* This case is for when the encoded fields are stored in a
* single avpkt but the hardware returns then separately. Unless
* we grab the second field before returning, we'll slip another
* frame in the pipeline and if that happens a lot, we're sunk.
* So we have to get that second field now.
* Interlaced mpeg2 and vc1 are examples of this.
*/
av_log(avctx, AV_LOG_VERBOSE, "Trying to get second field.\n");
while (1) {
usleep(priv->decode_wait);
ret = DtsGetDriverStatus(dev, &decoder_status);
if (ret == BC_STS_SUCCESS &&
decoder_status.ReadyListCount > 0) {
CrystalHD: Improve detection of h.264 content. As previously discussed, the CrystalHD hardware returns exceptionally useless information about interlaced h.264 content - to the extent that it's not possible to distinguish most MBAFF and PAFF content until it's too late. In an attempt to compensate for this, I'm introducing two mechanisms: 1) Peeking at the picture number of the next picture The hardware provides a capability to peek the next picture number. If it is the same as the current picture number, then we are clearly dealing with two fields and not a frame or fieldpair. If this always worked, it would be all we need, but it's not guaranteed to work. Sometimes, the next picture may not be decoded sufficiently for the number to be known; alternately, a corruption in the stream may cause the hardware to refuse to return the number even if the next intact frame is decoded. In either case, the query will return 0. If we are unable to peek the next picture number, we assume that the picture is a frame/fieldpair and return it accordingly. If that turns out to be incorrect, we discard the second field, and the user has to live with the glitch. In testing, false detection can occur for the first couple of seconds, and then the pipeline stabalizes and we get correct detection. 2) Use the h264_parser to detect when individual input fields have been combined into an output fieldpair. I have multiple PAFF samples where this behaviour is detected. The peeking mechanism described above will correctly detect that the output is a fieldpair, but we need to know what the input type was to ensure pipeline stability (only return one output frame per input frame). If we find ourselves with an output fieldpair, yet the input picture type was a field, as reported by the parser, then we are dealing with this case, and can make sure not to return anything on the next decode() call. Taken together, these allow us to remove the hard-coded hacks for different h.264 types, and we can clearly describe the conditions under which we can trust the hardware's claim that content is interlaced. Signed-off-by: Philip Langdale <philipl@overt.org>
2011-03-26 19:34:20 +02:00
rec_ret = receive_frame(avctx, data, data_size);
if ((rec_ret == RET_OK && *data_size > 0) ||
rec_ret == RET_ERROR)
break;
}
}
av_log(avctx, AV_LOG_VERBOSE, "CrystalHD: Got second field.\n");
} else if (rec_ret == RET_SKIP_NEXT_COPY) {
/*
* Two input packets got turned into a field pair. Gawd.
*/
av_log(avctx, AV_LOG_VERBOSE,
"Don't output on next decode call.\n");
priv->skip_next_output = 1;
}
/*
* If rec_ret == RET_COPY_AGAIN, that means that either we just handled
* a FMT_CHANGE event and need to go around again for the actual frame,
* we got a busy status and need to try again, or we're dealing with
* packed b-frames, where the hardware strangely returns the packed
* p-frame twice. We choose to keep the second copy as it carries the
* valid pts.
*/
} while (rec_ret == RET_COPY_AGAIN);
usleep(priv->decode_wait);
return len;
}
#if CONFIG_H264_CRYSTALHD_DECODER
static AVClass h264_class = {
"h264_crystalhd",
av_default_item_name,
options,
LIBAVUTIL_VERSION_INT,
};
AVCodec ff_h264_crystalhd_decoder = {
.name = "h264_crystalhd",
.type = AVMEDIA_TYPE_VIDEO,
.id = CODEC_ID_H264,
.priv_data_size = sizeof(CHDContext),
.init = init,
.close = uninit,
.decode = decode,
.capabilities = CODEC_CAP_DR1 | CODEC_CAP_DELAY,
.flush = flush,
.long_name = NULL_IF_CONFIG_SMALL("H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10 (CrystalHD acceleration)"),
.pix_fmts = (const enum PixelFormat[]){PIX_FMT_YUYV422, PIX_FMT_NONE},
.priv_class = &h264_class,
};
#endif
#if CONFIG_MPEG2_CRYSTALHD_DECODER
static AVClass mpeg2_class = {
"mpeg2_crystalhd",
av_default_item_name,
options,
LIBAVUTIL_VERSION_INT,
};
AVCodec ff_mpeg2_crystalhd_decoder = {
.name = "mpeg2_crystalhd",
.type = AVMEDIA_TYPE_VIDEO,
.id = CODEC_ID_MPEG2VIDEO,
.priv_data_size = sizeof(CHDContext),
.init = init,
.close = uninit,
.decode = decode,
.capabilities = CODEC_CAP_DR1 | CODEC_CAP_DELAY,
.flush = flush,
.long_name = NULL_IF_CONFIG_SMALL("MPEG-2 Video (CrystalHD acceleration)"),
.pix_fmts = (const enum PixelFormat[]){PIX_FMT_YUYV422, PIX_FMT_NONE},
.priv_class = &mpeg2_class,
};
#endif
#if CONFIG_MPEG4_CRYSTALHD_DECODER
static AVClass mpeg4_class = {
"mpeg4_crystalhd",
av_default_item_name,
options,
LIBAVUTIL_VERSION_INT,
};
AVCodec ff_mpeg4_crystalhd_decoder = {
.name = "mpeg4_crystalhd",
.type = AVMEDIA_TYPE_VIDEO,
.id = CODEC_ID_MPEG4,
.priv_data_size = sizeof(CHDContext),
.init = init,
.close = uninit,
.decode = decode,
.capabilities = CODEC_CAP_DR1 | CODEC_CAP_DELAY,
.flush = flush,
.long_name = NULL_IF_CONFIG_SMALL("MPEG-4 Part 2 (CrystalHD acceleration)"),
.pix_fmts = (const enum PixelFormat[]){PIX_FMT_YUYV422, PIX_FMT_NONE},
.priv_class = &mpeg4_class,
};
#endif
#if CONFIG_MSMPEG4_CRYSTALHD_DECODER
static AVClass msmpeg4_class = {
"msmpeg4_crystalhd",
av_default_item_name,
options,
LIBAVUTIL_VERSION_INT,
};
AVCodec ff_msmpeg4_crystalhd_decoder = {
.name = "msmpeg4_crystalhd",
.type = AVMEDIA_TYPE_VIDEO,
.id = CODEC_ID_MSMPEG4V3,
.priv_data_size = sizeof(CHDContext),
.init = init,
.close = uninit,
.decode = decode,
.capabilities = CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_EXPERIMENTAL,
.flush = flush,
.long_name = NULL_IF_CONFIG_SMALL("MPEG-4 Part 2 Microsoft variant version 3 (CrystalHD acceleration)"),
.pix_fmts = (const enum PixelFormat[]){PIX_FMT_YUYV422, PIX_FMT_NONE},
.priv_class = &msmpeg4_class,
};
#endif
#if CONFIG_VC1_CRYSTALHD_DECODER
static AVClass vc1_class = {
"vc1_crystalhd",
av_default_item_name,
options,
LIBAVUTIL_VERSION_INT,
};
AVCodec ff_vc1_crystalhd_decoder = {
.name = "vc1_crystalhd",
.type = AVMEDIA_TYPE_VIDEO,
.id = CODEC_ID_VC1,
.priv_data_size = sizeof(CHDContext),
.init = init,
.close = uninit,
.decode = decode,
.capabilities = CODEC_CAP_DR1 | CODEC_CAP_DELAY,
.flush = flush,
.long_name = NULL_IF_CONFIG_SMALL("SMPTE VC-1 (CrystalHD acceleration)"),
.pix_fmts = (const enum PixelFormat[]){PIX_FMT_YUYV422, PIX_FMT_NONE},
.priv_class = &vc1_class,
};
#endif
#if CONFIG_WMV3_CRYSTALHD_DECODER
static AVClass wmv3_class = {
"wmv3_crystalhd",
av_default_item_name,
options,
LIBAVUTIL_VERSION_INT,
};
AVCodec ff_wmv3_crystalhd_decoder = {
.name = "wmv3_crystalhd",
.type = AVMEDIA_TYPE_VIDEO,
.id = CODEC_ID_WMV3,
.priv_data_size = sizeof(CHDContext),
.init = init,
.close = uninit,
.decode = decode,
.capabilities = CODEC_CAP_DR1 | CODEC_CAP_DELAY,
.flush = flush,
.long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9 (CrystalHD acceleration)"),
.pix_fmts = (const enum PixelFormat[]){PIX_FMT_YUYV422, PIX_FMT_NONE},
.priv_class = &wmv3_class,
};
#endif