[FFmpeg-devel] [PATCH] avcodec/av1dec: fix loading PrevGmParams for frames with primary_ref_frame none

James Almer jamrial at gmail.com
Tue Oct 27 23:07:49 EET 2020


On 10/27/2020 6:04 PM, Mark Thompson wrote:
> On 27/10/2020 21:01, James Almer wrote:
>> On 10/27/2020 5:57 PM, Mark Thompson wrote:
>>> On 17/10/2020 20:30, James Almer wrote:
>>>> Signed-off-by: James Almer <jamrial at gmail.com>
>>>> ---
>>>>    libavcodec/av1dec.c | 9 +++++++--
>>>>    1 file changed, 7 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/libavcodec/av1dec.c b/libavcodec/av1dec.c
>>>> index 54aeba1812..04aaf5d148 100644
>>>> --- a/libavcodec/av1dec.c
>>>> +++ b/libavcodec/av1dec.c
>>>> @@ -109,13 +109,18 @@ static void read_global_param(AV1DecContext *s,
>>>> int type, int ref, int idx)
>>>>    {
>>>>        uint8_t primary_frame, prev_frame;
>>>>        uint32_t abs_bits, prec_bits, round, prec_diff, sub, mx;
>>>> -    int32_t r;
>>>> +    int32_t r, prev_gm_param;
>>>>          primary_frame = s->raw_frame_header->primary_ref_frame;
>>>>        prev_frame = s->raw_frame_header->ref_frame_idx[primary_frame];
>>>>        abs_bits = AV1_GM_ABS_ALPHA_BITS;
>>>>        prec_bits = AV1_GM_ALPHA_PREC_BITS;
>>>>    +    if (s->raw_frame_header->primary_ref_frame ==
>>>> AV1_PRIMARY_REF_NONE)
>>>> +        prev_gm_param = s->cur_frame.gm_params[ref][idx];
>>>
>>> To clarify that I'm reading the standard correctly here, this is the
>>> value set for PrevGmParams by setup_past_independance(), which then
>>> matches the default for gm_params at the top of global_motion_params()
>>> so you've reused it here?
>>
>> Correct. Assuming s->cur_frame.gm_params will contain the default values
>> when primary_ref_frame == AV1_PRIMARY_REF_NONE (as it happens in
>> global_motion_params()) simplifies the code quite a bit. Otherwise I'd
>> have to add some array with said defaults somewhere.
> 
> Do you mind adding a little comment saying that?  It took me quite a few
> minutes of staring at it thinking it was wrong to realise what was going
> on.

Sure. And sorry.

> 
>>>
>>>> +    else
>>>> +        prev_gm_param = s->ref[prev_frame].gm_params[ref][idx];
>>>> +
>>>>        if (idx < 2) {
>>>>            if (type == AV1_WARP_MODEL_TRANSLATION) {
>>>>                abs_bits = AV1_GM_ABS_TRANS_ONLY_BITS -
>>>> @@ -131,7 +136,7 @@ static void read_global_param(AV1DecContext *s,
>>>> int type, int ref, int idx)
>>>>        prec_diff = AV1_WARPEDMODEL_PREC_BITS - prec_bits;
>>>>        sub = (idx % 3) == 2 ? (1 << prec_bits) : 0;
>>>>        mx = 1 << abs_bits;
>>>> -    r = (s->ref[prev_frame].gm_params[ref][idx] >> prec_diff) - sub;
>>>> +    r = (prev_gm_param >> prec_diff) - sub;
>>>>          s->cur_frame.gm_params[ref][idx] =
>>>>          
>>>> (decode_signed_subexp_with_ref(s->raw_frame_header->gm_params[ref][idx],
>>>>
>>>>
> 
> LGTM with that.
> 
> Thanks,

Will apply, thanks.

> 
> - Mark
> _______________________________________________
> 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".



More information about the ffmpeg-devel mailing list