[FFmpeg-devel] [PATCH 3/3] avcodec/h264_slice: use ff_h264_replace_picture when updating thread contexts

[James Almer]

On 8/8/2021 8:22 PM, Andreas Rheinhardt wrote:
> James Almer:
>> Signed-off-by: James Almer <jamrial at gmail.com>
>> ---
>>   libavcodec/h264_slice.c | 14 +++++---------
>>   1 file changed, 5 insertions(+), 9 deletions(-)
>>
>> diff --git a/libavcodec/h264_slice.c b/libavcodec/h264_slice.c
>> index a31e804620..0d7107d455 100644
>> --- a/libavcodec/h264_slice.c
>> +++ b/libavcodec/h264_slice.c
>> @@ -379,19 +379,15 @@ int ff_h264_update_thread_context(AVCodecContext *dst,
>>       h->droppable            = h1->droppable;
>>   
>>       for (i = 0; i < H264_MAX_PICTURE_COUNT; i++) {
>> -        ff_h264_unref_picture(h, &h->DPB[i]);
>> -        if (h1->DPB[i].f->buf[0] &&
>> -            (ret = ff_h264_ref_picture(h, &h->DPB[i], &h1->DPB[i])) < 0)
>> +        ret = ff_h264_replace_picture(h, &h->DPB[i], &h1->DPB[i]);
>> +        if (ret < 0)
>>               return ret;
>>       }
>>   
>>       h->cur_pic_ptr = REBASE_PICTURE(h1->cur_pic_ptr, h, h1);
>> -    ff_h264_unref_picture(h, &h->cur_pic);
>> -    if (h1->cur_pic.f->buf[0]) {
>> -        ret = ff_h264_ref_picture(h, &h->cur_pic, &h1->cur_pic);
>> -        if (ret < 0)
>> -            return ret;
>> -    }
>> +    ret = ff_h264_replace_picture(h, &h->cur_pic, &h1->cur_pic);
>> +    if (ret < 0)
>> +        return ret;
>>   
>>       h->enable_er       = h1->enable_er;
>>       h->workaround_bugs = h1->workaround_bugs;
>>
> I like what you are doing here; I once had the same idea, but IIRC it
> was way uglier.
> Can you share some numbers on how many allocations are avoided by this?

I did not look into what specific buffers within H264Picture were not 
allocated (unref + ref) after this patch, but i did check what calls to 
ff_h264_replace_picture() had to do it for at least one of them, and in 
a -threads 4 scenario for a standard 1080p high profile sample cur_pic 
pretty much always generated allocations, but only about three out of 
all DPB pictures (about sixteen) did it.

Considering the above, it would be nice if we could do the same for the 
AVFrame too, but when i looked at adding an av_frame_replace() it looked 
like it wouldn't be trivial.

> (I guess that avoiding the allocation will be the normal case for at
> least the PPS buffer.) How many would it be with this patch:
> https://ffmpeg.org/pipermail/ffmpeg-devel/2020-November/272508.html?
> 
> - Andreas
> _______________________________________________
> 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".
>