Re: [PATCH] media: hantro: HEVC: Fix tile info buffer value computation

[Benjamin Gaignard]

Le 27/04/2022 à 16:22, Dan Carpenter a écrit :
> On Wed, Apr 27, 2022 at 03:55:17PM +0200, Benjamin Gaignard wrote:
> > Use pps->column_width_minus1[j] + 1 as value for the tile info buffer
> > instead of pps->column_width_minus1[j + 1].
> > The patch fix DBLK_E_VIXS_2, DBLK_F_VIXS_2, DBLK_G_VIXS_2,
> > SAO_B_MediaTek_5, TILES_A_Cisco_2 and TILES_B_Cisco_1 tests in fluster.
> >
> > Signed-off-by: Benjamin Gaignard <benjamin.gaignard@xxxxxxxxxxxxx>
> Fixes: cb5dd5a0fa51 ("media: hantro: Introduce G2/HEVC decoder")
>
> > ---
> >   drivers/staging/media/hantro/hantro_g2_hevc_dec.c | 2 +-
> >   1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/staging/media/hantro/hantro_g2_hevc_dec.c b/drivers/staging/media/hantro/hantro_g2_hevc_dec.c
> > index bb512389c1a5..ffeb2fbeefd2 100644
> > --- a/drivers/staging/media/hantro/hantro_g2_hevc_dec.c
> > +++ b/drivers/staging/media/hantro/hantro_g2_hevc_dec.c
> > @@ -74,7 +74,7 @@ static void prepare_tile_info_buffer(struct hantro_ctx *ctx)
> >   					no_chroma = 1;
> >   				for (j = 0, tmp_w = 0; j < num_tile_cols - 1; j++) {
>                                                                          ^^^^
> Is this "- 1" still required?  I would have assumed the "- 1" here was
> connected with the [j + 1].  What is stored in the last element of the
> ->column_width_minus1 array?

No "- 1" comes from the definition of the field in H.265 which is pps->num_tile_columns_minus1
We add 1 to fill the registers so we need the "- 1" in the loop.
Data ordering isn't documented anywhere so that part is directly inspired by
Hantro/G2 proprietary stack.

Regards,
Benjamin

> >   					tmp_w += pps->column_width_minus1[j] + 1;
> > -					*p++ = pps->column_width_minus1[j + 1];
>                                                                          ^^^^^
>
> > +					*p++ = pps->column_width_minus1[j] + 1;
> >   					*p++ = h;
> >   					if (i == 0 && h == 1 && ctb_size == 16)
> >   						no_chroma = 1;
> regards,
> dan carpenter