Now that the support for hor/vert downsampled planar formats is in
place we can add support for such formats to the TPG.
This patch adds support for:
V4L2_PIX_FMT_YUV420M
V4L2_PIX_FMT_YVU420M
V4L2_PIX_FMT_YUV420
V4L2_PIX_FMT_YVU420
V4L2_PIX_FMT_YUV422P
V4L2_PIX_FMT_NV16
V4L2_PIX_FMT_NV61
V4L2_PIX_FMT_NV12
V4L2_PIX_FMT_NV21
V4L2_PIX_FMT_NV12P
V4L2_PIX_FMT_NV21P
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Implement horizontal and vertical downsampling when filling in the
plane. The TPG is now ready to support such formats.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
This will just skip lines/pixels since color fidelity is not quite
as important here as it is with the test patterns themselves.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
gen_twopix() is always called twice: once for the first and once for
the second pixel. Improve the code to properly average the two if the
format requires horizontal downsampling.
This is necessary for patterns like 1x1 red/blue checkers.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
When dealing with vertical downsampling two successive lines have to be
averaged. In the case of the test pattern generator that only happens
if the two lines are using different patterns. So precalculate the average
between two pattern lines: one of pattern P and one of pattern P + 1.
That way there is no need to do any on-the-fly downsampling: it's all done
in the precalculate phase.
This patch also implements horizontal downsampling in the precalculate phase.
The only thing that needs to be done is to half the width since the actual
downsampling happens when two pixels at a time are generated.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
This is step one of supporting horizontal and vertical downsampling.
This just adds support for the h/vdownsampling fields and it increases
the maximum number of planes to 3.
Currently none of the planar formats need horizontal or vertical
downsampling, so this change has no effect at the moment.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Add helpers functions to determine the line widths and image sizes
for planar formats that are stores in a single buffer.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Add a new field that contains the number of buffers. This may be
less than the number of planes in case multiple planes are combined
into one buffer.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
The grayscale values are still full range sRGB, so don't add the
limited range offset.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Add a 2x2 checker patterns and 1x1 and 2x2 red/blue checker patterns.
Useful for testing 4:2:2 and 4:2:0 formats.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
The successive TOP/BOTTOM fields did not move as they should: only
every other field actually changed position.
The cause was that the tpg needs to know if it is dealing with a
FIELD_ALTERNATE case since that requires slightly different handling.
So tell the TPG whether or not the field setting is for the ALTERNATE
case or not.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
According to the standard the R'G'B' BT.2020 colorspace is limited
range, not full range. Fix this.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
When converting to or from Y'CbCr and R'G'B' the Y offset depends
on the quantization range: it's 0 for full and 16 for limited range.
But in the code it was hardcoded to 16. This messed up the brightness
of the generated pattern.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Add support for the new AdobeRGB and BT.2020 colorspaces. Also support
explicit Y'CbCr and quantization settings.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
The random_line buffer must be twice the maximum width, but it only allocated
the maximum width, so it was only half the size it needed to be.
Surprisingly I never saw the kernel fail on this, but the same TPG code used in
qv4l2 crashed and valgrind helped me track this bug down.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Replaced enum v4l2_field by u32, just as the prototype in the
header has.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>
The test patterns for video capture are generated by this code. All patterns
are precalculated taking into account colorspace information, pixel and video
aspect ratios and scaling information.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>
