[FFmpeg-devel] [PATCH] avformat/adxdec: demux multiple blocks at once

[Andreas Rheinhardt]

Paul B Mahol:
> On Fri, Sep 18, 2020 at 12:16:18PM +0200, Andreas Rheinhardt wrote:
>> Paul B Mahol:
>>> Improves decoding speed by 24x
>>>
>>> Signed-off-by: Paul B Mahol <onemda at gmail.com>
>>> ---
>>>  libavformat/adxdec.c | 23 +++++++++++++++--------
>>>  1 file changed, 15 insertions(+), 8 deletions(-)
>>>
>>> diff --git a/libavformat/adxdec.c b/libavformat/adxdec.c
>>> index ccd5049acd..0e4c251cfc 100644
>>> --- a/libavformat/adxdec.c
>>> +++ b/libavformat/adxdec.c
>>> @@ -53,6 +53,9 @@ static int adx_read_packet(AVFormatContext *s, AVPacket *pkt)
>>>      AVCodecParameters *par = s->streams[0]->codecpar;
>>>      int ret, size;
>>>  
>>> +    if (avio_feof(s->pb))
>>> +        return AVERROR_EOF;
>>> +
>>>      if (par->channels <= 0) {
>>>          av_log(s, AV_LOG_ERROR, "invalid number of channels %d\n", par->channels);
>>>          return AVERROR_INVALIDDATA;
>>> @@ -63,16 +66,20 @@ static int adx_read_packet(AVFormatContext *s, AVPacket *pkt)
>>>      pkt->pos = avio_tell(s->pb);
>>>      pkt->stream_index = 0;
>>>  
>>> -    ret = av_get_packet(s->pb, pkt, size);
>>> -    if (ret != size) {
>>> -        return ret < 0 ? ret : AVERROR(EIO);
>>> -    }
>>> -    if (AV_RB16(pkt->data) & 0x8000) {
>>> -        return AVERROR_EOF;
>>> +    ret = av_get_packet(s->pb, pkt, size * 128);
>>> +    if (ret < 0)
>>> +        return ret;
>>> +    if ((ret % size) && ret >= size) {
>>
>> So if ret < size you don't set the corrupt flag. Why?
> 
> Because data, that is not gonna be used at all, is discarded, instead of
> pointlessly being passed to decoder and there errored out.
> This happens at every single eof.
> 

But you don't return an error if the return value is in the range
0..size - 1; you return the packet with the incomplete and apparently
useless packet.

>>
>>> +        size = ret - (ret % size);
>>> +        av_shrink_packet(pkt, size);
>>> +        pkt->flags &= ~AV_PKT_FLAG_CORRUPT;
>>> +    } else {
>>> +        size = ret;
>>>      }
>>> +
>>>      pkt->size     = size;
>>
>> This line makes no sense any more: If the first branch above is taken,
>> av_shrink_packet() will already set the size; in the other branch,
>> av_get_packet() already did.
> 
> Removed that line locally.
> 
>>
>>> -    pkt->duration = 1;
>>> -    pkt->pts      = (pkt->pos - c->header_size) / size;
>>> +    pkt->duration = size / (BLOCK_SIZE * par->channels);
>>> +    pkt->pts      = (pkt->pos - c->header_size) / (BLOCK_SIZE * par->channels);
>>>  
>>>      return 0;
>>>  }
>>>
>>
>> _______________________________________________
>> ffmpeg-devel mailing list
>> ffmpeg-devel at ffmpeg.org
>> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>>
>> To unsubscribe, visit link above, or email
>> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> 
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".
>