Re: [PATCH 4/9] media: vsp1: Add support for missing 32-bit RGB formats

From: Jacopo Mondi <jacopo@jmondi.org>
To: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Cc: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>,
	dri-devel@lists.freedesktop.org, linux-media@vger.kernel.org,
	linux-renesas-soc@vger.kernel.org,
	Maxime Ripard <maxime.ripard@bootlin.com>
Subject: Re: [PATCH 4/9] media: vsp1: Add support for missing 32-bit RGB formats
Date: Tue, 23 Apr 2019 15:21:53 +0200	[thread overview]
Message-ID: <20190423132153.xfvy4dn3cflzotuf@uno.localdomain> (raw)
In-Reply-To: <20190418060645.GC4806@pendragon.ideasonboard.com>

[-- Attachment #1: Type: text/plain, Size: 4074 bytes --]

Hi Laurent,

On Thu, Apr 18, 2019 at 09:06:45AM +0300, Laurent Pinchart wrote:
> Hi Jacopo,
>
> On Thu, Apr 04, 2019 at 07:39:05PM +0200, Jacopo Mondi wrote:
> > HI Laurent,
> >   if you help me out understanding the bit swapping procedure in VSP
> > I would be more confident in saying I actually reviewed the series.
>
> I'll try my best :-)
>

Thanks!

> > On Thu, Mar 28, 2019 at 09:07:18AM +0200, Laurent Pinchart wrote:
> > > Add support for the V4L2_PIX_FMT_BGRA32, V4L2_PIX_FMT_BGRX32,
> > > V4L2_PIX_FMT_RGBA32 and V4L2_PIX_FMT_RGBX32 formats to the VSP driver.
> > >
> > > Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
> > > ---
> > >  drivers/media/platform/vsp1/vsp1_pipe.c | 14 ++++++++++++++
> > >  1 file changed, 14 insertions(+)
> > >
> > > diff --git a/drivers/media/platform/vsp1/vsp1_pipe.c b/drivers/media/platform/vsp1/vsp1_pipe.c
> > > index 54ff539ffea0..9f08d85e89d7 100644
> > > --- a/drivers/media/platform/vsp1/vsp1_pipe.c
> > > +++ b/drivers/media/platform/vsp1/vsp1_pipe.c
> > > @@ -68,6 +68,20 @@ static const struct vsp1_format_info vsp1_video_formats[] = {
> > >  	{ V4L2_PIX_FMT_XBGR32, MEDIA_BUS_FMT_ARGB8888_1X32,
> > >  	  VI6_FMT_ARGB_8888, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS,
> > >  	  1, { 32, 0, 0 }, false, false, 1, 1, false },
> > > +	{ V4L2_PIX_FMT_BGRA32, MEDIA_BUS_FMT_ARGB8888_1X32,
> > > +	  VI6_FMT_RGBA_8888, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS,
> > > +	  1, { 32, 0, 0 }, false, false, 1, 1, true },
> > > +	{ V4L2_PIX_FMT_BGRX32, MEDIA_BUS_FMT_ARGB8888_1X32,
> > > +	  VI6_FMT_RGBA_8888, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS,
> > > +	  1, { 32, 0, 0 }, false, false, 1, 1, false },
> > > +	{ V4L2_PIX_FMT_RGBA32, MEDIA_BUS_FMT_ARGB8888_1X32,
> > > +	  VI6_FMT_RGBA_8888, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > +	  VI6_RPF_DSWAP_P_WDS | VI6_RPF_DSWAP_P_BTS,
> > > +	  1, { 32, 0, 0 }, false, false, 1, 1, true },
> > > +	{ V4L2_PIX_FMT_RGBX32, MEDIA_BUS_FMT_ARGB8888_1X32,
> > > +	  VI6_FMT_RGBA_8888, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > +	  VI6_RPF_DSWAP_P_WDS | VI6_RPF_DSWAP_P_BTS,
> > > +	  1, { 32, 0, 0 }, false, false, 1, 1, false },
> >
> >
> > So, this is how I interpret the above lines for format BGRA32 (the
> > same applies to RGBA32 fwiw)
> >
> >         Bytes:          B3      B2      B1      B0
> > Input: ARGB8888_1X32:   A       R       G       B
>
> Note that, at the bus level, there's no byte ordering. Data is
> transmitted in parallel on a 32-bit bus (at least as far as I know). But
> this doesn't matter much anyway.
>

Indeed, but I should not use this as the description of the input
format but VI6_FMT_RGBA_8888, as you pointed out, which reads indeed
as RGBA (out of curiosity, what is the mbus code used for then?)

> > Out:   BGRA32           R       G       B       A
> >
> > If I apply LLS and LWS on the "Input Row", according to the table I get
> >
> >  Position in Table 32.10:    0  1  2  3
> >  IN Components ordering:     A  R  G  B  (ARGB8888_1X32
> >  Position in Table 32.10:    3  2  1  0  (with LLS and LWS applied)
> >  OUT Components ordering:    B  G  R  A
> >
> > Which to me is the ordering of components of the format named
> > V4L2_PIX_FMT_ARGB32 and not BGRA32 which is instead, as reported RGBA
> > (GREAT naming here!)
> >
> > What am I doing wrong?
>
> You need to take entry VI6_FMT_RGBA_8888 (0x14) in table 32.10, which
> reads R G B A. Applying LLS and LWS but not WDS and BTS swaps the bytes
> to A B G R (see table 32.13), so the pixel is written to memory in the A
> B G R order, corresponding to V4L2_PIX_FMT_BGRA32 (I agree with you
> about the insane naming for the 4CC, but that's not my fault :-().

Thanks, now that I applied flags to the correct input format, the
patch seems correct to me.

Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>

Thanks
   j

>
> > >  	{ V4L2_PIX_FMT_ARGB32, MEDIA_BUS_FMT_ARGB8888_1X32,
> > >  	  VI6_FMT_ARGB_8888, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > >  	  VI6_RPF_DSWAP_P_WDS | VI6_RPF_DSWAP_P_BTS,
>
> --
> Regards,
>
> Laurent Pinchart

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]