2018-08-01 19:13:53

by Steve Longerbeam

Subject: [PATCH v3 04/14] gpu: ipu-v3: Fix U/V offset macros for planar 4:2:0

The U and V offset macros for planar 4:2:0 (U_OFFSET, V_OFFSET, and
UV_OFFSET), are not correct. The height component to the offset was
calculated as:

(pix->width * y / 4)

But this does not produce correct offsets for odd values of y (luma
line #). The luma line # must be decimated by two to produce the
correct U/V line #, so the correct formula is:

(pix->width * (y / 2) / 2)

Signed-off-by: Steve Longerbeam <[email protected]>
---
drivers/gpu/ipu-v3/ipu-cpmem.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/ipu-v3/ipu-cpmem.c b/drivers/gpu/ipu-v3/ipu-cpmem.c
index 9f2d9ec..e68e473 100644
--- a/drivers/gpu/ipu-v3/ipu-cpmem.c
+++ b/drivers/gpu/ipu-v3/ipu-cpmem.c
@@ -530,17 +530,17 @@ static const struct ipu_rgb def_bgra_16 = {

#define Y_OFFSET(pix, x, y) ((x) + pix->width * (y))
#define U_OFFSET(pix, x, y) ((pix->width * pix->height) + \
- (pix->width * (y) / 4) + (x) / 2)
+ (pix->width * ((y) / 2) / 2) + (x) / 2)
#define V_OFFSET(pix, x, y) ((pix->width * pix->height) + \
(pix->width * pix->height / 4) + \
- (pix->width * (y) / 4) + (x) / 2)
+ (pix->width * ((y) / 2) / 2) + (x) / 2)
#define U2_OFFSET(pix, x, y) ((pix->width * pix->height) + \
(pix->width * (y) / 2) + (x) / 2)
#define V2_OFFSET(pix, x, y) ((pix->width * pix->height) + \
(pix->width * pix->height / 2) + \
(pix->width * (y) / 2) + (x) / 2)
#define UV_OFFSET(pix, x, y) ((pix->width * pix->height) + \
- (pix->width * (y) / 2) + (x))
+ (pix->width * ((y) / 2)) + (x))
#define UV2_OFFSET(pix, x, y) ((pix->width * pix->height) + \
(pix->width * y) + (x))

--
2.7.4

2018-08-02 09:53:10

by Philipp Zabel

Subject: Re: [PATCH v3 04/14] gpu: ipu-v3: Fix U/V offset macros for planar 4:2:0

On Wed, 2018-08-01 at 12:12 -0700, Steve Longerbeam wrote:
> The U and V offset macros for planar 4:2:0 (U_OFFSET, V_OFFSET, and
> UV_OFFSET), are not correct. The height component to the offset was
> calculated as:
>
> (pix->width * y / 4)
>
> But this does not produce correct offsets for odd values of y (luma
> line #). The luma line # must be decimated by two to produce the
> correct U/V line #, so the correct formula is:
>
> (pix->width * (y / 2) / 2)
>
> Signed-off-by: Steve Longerbeam <[email protected]>

Thank you this patch is applied to imx-drm/fixes now.

regards
Philipp