[Libva] [PATCH][Libva-intel-driver] i965_DeriveImage() support JPEG color formats

[Gwenole Beauchesne]

2014-04-17 2:55 GMT+02:00 Zhao Yakui <yakui.zhao at intel.com>:
> On Wed, 2014-04-16 at 06:23 -0600, Gwenole Beauchesne wrote:
>> 2014-04-15 17:48 GMT-07:00 Xiang, Haihao <haihao.xiang at intel.com>:
>> > On Tue, 2014-04-15 at 07:27 -0700, Gwenole Beauchesne wrote:
>> >> Hi,
>> >>
>> >> 2014-04-14 1:17 GMT-07:00 Zhong Li <zhong.li at intel.com>:
>> >> > Signed-off-by: Zhong Li <zhong.li at intel.com>
>> >> > ---
>> >> >  src/i965_drv_video.c |    6 ++++++
>> >> >  1 file changed, 6 insertions(+)
>> >> >
>> >> > diff --git a/src/i965_drv_video.c b/src/i965_drv_video.c
>> >> > index cc61ad3..f94d429 100755
>> >> > --- a/src/i965_drv_video.c
>> >> > +++ b/src/i965_drv_video.c
>> >> > @@ -3286,6 +3286,10 @@ VAStatus i965_DeriveImage(VADriverContextP ctx,
>> >> >
>> >> >      case VA_FOURCC_I420:
>> >> >      case VA_FOURCC_422H:
>> >> > +    case VA_FOURCC_IMC3:
>> >> > +    case VA_FOURCC_444P:
>> >> > +    case VA_FOURCC_422V:
>> >> > +    case VA_FOURCC_411P:
>> >> >          image->num_planes = 3;
>> >> >          image->pitches[0] = w_pitch; /* Y */
>> >> >          image->offsets[0] = 0;
>> >>
>> >> IMC3 fourcc is totally useless, change that to be I420.
>> >
>> > IMC3 is required for JPEG decoding and U/V plane + garbage is in full
>> > pitch.
>>
>> And? IMC3 = I420 with stride for each plane that is the same, and that
>> is specified in the VAImage struct. So, what is the point? This needs
>> to be changed. That's all an application needs to know. i.e. move up
>> VA_FOURCC_IMC3 to the top, make image->fourcc = VA_FOURCC_I420, and
>> fall-through in the rest...
>
> The IMC3 and I420 have the similar plane organization(Y/U/V).
>
> Is there any benefit that the I965_deriveImage exports image->fourcc as
> the VA_FOURCC_I420 when the surface is VA_FOURCC_IMC3?

Yes, every single application out there directly knows about I420, not IMC3.

> Currently the fourcc of image format for the derived image is the same
> as the surface fourcc.
> If we change it, it seems very strange for the IMC3.

It's not strange, but useful and practical.

>> >> >          image->pitches[2] = obj_surface->cb_cr_pitch; /* V */
>> >> >          image->offsets[2] = w_pitch * obj_surface->y_cr_offset;
>> >> >          break;
>> >> > +
>> >> >      case VA_FOURCC_YUY2:
>> >> >      case VA_FOURCC_UYVY:
>> >> > +    case VA_FOURCC_Y800:
>> >> >          image->num_planes = 1;
>> >> >          image->pitches[0] = obj_surface->width; /* Y, width is aligned already */
>> >> >          image->offsets[0] = 0;
>> >> > --
>> >> > 1.7.9.5
>> >> >
>> >> > _______________________________________________
>> >> > Libva mailing list
>> >> > Libva at lists.freedesktop.org
>> >> > http://lists.freedesktop.org/mailman/listinfo/libva
>> >> _______________________________________________
>> >> Libva mailing list
>> >> Libva at lists.freedesktop.org
>> >> http://lists.freedesktop.org/mailman/listinfo/libva
>> >
>> >
>> _______________________________________________
>> Libva mailing list
>> Libva at lists.freedesktop.org
>> http://lists.freedesktop.org/mailman/listinfo/libva
>
>