[FFmpeg-devel] [PATCH] mdec: use correctly permutated quant matrix for dequantization.

[Ronald S. Bultje]

Hi,

On Tue, Jun 20, 2017 at 12:04 PM, James Darnley <jdarnley at obe.tv> wrote:

> On 2017-06-20 14:47, Ronald S. Bultje wrote:
> > This allows using non-simple (e.g. simplemmx) IDCT implementations.
> > The result is not bitexact (which is why the fate test continues to
> > use -idct simple), but the PSNR between C/MMX goes from ~35 to ~90.
> > ---
> >  libavcodec/mdec.c | 14 ++++++++++----
> >  1 file changed, 10 insertions(+), 4 deletions(-)
> >
> > diff --git a/libavcodec/mdec.c b/libavcodec/mdec.c
> > index 97bfebb..59658b3 100644
> > --- a/libavcodec/mdec.c
> > +++ b/libavcodec/mdec.c
> > @@ -49,6 +49,7 @@ typedef struct MDECContext {
> >      int mb_height;
> >      int mb_x, mb_y;
> >      DECLARE_ALIGNED(16, int16_t, block)[6][64];
> > +    DECLARE_ALIGNED(16, uint16_t, quant_matrix)[64];
> >      uint8_t *bitstream_buffer;
> >      unsigned int bitstream_buffer_size;
> >      int block_last_index[6];
> > @@ -61,7 +62,7 @@ static inline int mdec_decode_block_intra(MDECContext
> *a, int16_t *block, int n)
> >      int component;
> >      RLTable *rl = &ff_rl_mpeg1;
> >      uint8_t * const scantable = a->scantable.permutated;
> > -    const uint16_t *quant_matrix = ff_mpeg1_default_intra_matrix;
> > +    const uint16_t *quant_matrix = a->quant_matrix;
> >      const int qscale = a->qscale;
> >
> >      /* DC coefficient */
> > @@ -212,9 +213,7 @@ static int decode_frame(AVCodecContext *avctx,
> >  static av_cold int decode_init(AVCodecContext *avctx)
> >  {
> >      MDECContext * const a = avctx->priv_data;
> > -
> > -    if (avctx->idct_algo == FF_IDCT_AUTO)
> > -        avctx->idct_algo = FF_IDCT_SIMPLE;
> > +    int i;
> >
> >      a->mb_width  = (avctx->coded_width  + 15) / 16;
> >      a->mb_height = (avctx->coded_height + 15) / 16;
> > @@ -231,6 +230,13 @@ static av_cold int decode_init(AVCodecContext
> *avctx)
> >      avctx->pix_fmt  = AV_PIX_FMT_YUVJ420P;
> >      avctx->color_range = AVCOL_RANGE_JPEG;
> >
> > +    /* init q matrix */
> > +    for (i = 0; i < 64; i++) {
> > +        int j = a->idsp.idct_permutation[i];
> > +
> > +        a->quant_matrix[j] = ff_mpeg1_default_intra_matrix[i];
> > +    }
> > +
> >      return 0;
> >  }
> >
> >
>
> That patch seems to work.  FATE doesn't complain when I apply on top of
> my others after dropping the one Michael complained about.  That is what
> it was supposed to do, right?


Yes, it's intended to be an alternative to that patch, based on Michael's
review/comments.

Ronald