[FFmpeg-devel] [PATCH] mjpegdec: consider chroma subsampling in size check

[Andreas Cadhalpun]

On 05.12.2015 04:02, Michael Niedermayer wrote:
> On Fri, Dec 04, 2015 at 03:14:21PM +0100, Andreas Cadhalpun wrote:
>> On 03.12.2015 15:48, Michael Niedermayer wrote:
>>> On Wed, Dec 02, 2015 at 10:00:13PM +0100, Andreas Cadhalpun wrote:
>>>> @@ -1293,14 +1296,16 @@ static int mjpeg_decode_scan(MJpegDecodeContext *s, int nb_components, int Ah,
>>>>                  v = s->v_scount[i];
>>>>                  x = 0;
>>>>                  y = 0;
>>>> +                h_shift = c ? chroma_h_shift: 0;
>>>> +                v_shift = c ? chroma_v_shift: 0;
>>>>                  for (j = 0; j < n; j++) {
>>>>                      block_offset = (((linesize[c] * (v * mb_y + y) * 8) +
>>>>                                       (h * mb_x + x) * 8 * bytes_per_pixel) >> s->avctx->lowres);
>>>>  
>>>>                      if (s->interlaced && s->bottom_field)
>>>>                          block_offset += linesize[c] >> 1;
>>>> -                    if (   8*(h * mb_x + x) < s->width
>>>> -                        && 8*(v * mb_y + y) < s->height) {
>>>> +                    if (   8*(h * mb_x + x) < (s->width  + (1 << h_shift) - 1) >> h_shift
>>>> +                        && 8*(v * mb_y + y) < (s->height + (1 << v_shift) - 1) >> v_shift) {
>>>
>>> please move the w/h computation out of the block loop
>>> it stays the same for a component and does not need to be
>>> recalculated
>>> theres a loop above that fills data[] that can probably be used to
>>> fill w/h arrays
>>
>> OK, but since there are only two possible values, I don't think filling
>> arrays is necessary. Attached is an updated patch.
> 
> couldnt there be a alpha plane too ?

Yes, fixed patch attached.
I also tested filling w/h arrays, but that turned out to be slightly slower.

Best regards,
Andreas

-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-mjpegdec-consider-chroma-subsampling-in-size-check.patch
Type: text/x-diff
Size: 2443 bytes
Desc: not available
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20151206/522daf85/attachment.patch>