[v1,6/6] pipeline: vimc: Don't hardcode scaling factor with recent kernels

19951 diff mbox series

Starting in kernel v5.16, the vimc driver stopped hardcoding the scaler
factor. Use this to lift constraints on the camera configuration, and in
particular on the exotic output size alignment to a multiple of 6. As a
result, vimc-based cameras can more easily match common display
resolutions.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
---
 src/libcamera/pipeline/vimc/vimc.cpp | 47 +++++++++++++++++++---------
 1 file changed, 33 insertions(+), 14 deletions(-)

Quoting Laurent Pinchart (2024-04-25 00:42:24)
> Starting in kernel v5.16, the vimc driver stopped hardcoding the scaler
> factor. Use this to lift constraints on the camera configuration, and in
> particular on the exotic output size alignment to a multiple of 6. As a
> result, vimc-based cameras can more easily match common display
> resolutions.

Now that's helpful!

> Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
> ---
>  src/libcamera/pipeline/vimc/vimc.cpp | 47 +++++++++++++++++++---------
>  1 file changed, 33 insertions(+), 14 deletions(-)
> 
> diff --git a/src/libcamera/pipeline/vimc/vimc.cpp b/src/libcamera/pipeline/vimc/vimc.cpp
> index 5e66ee1d26c1..be7a6733472b 100644
> --- a/src/libcamera/pipeline/vimc/vimc.cpp
> +++ b/src/libcamera/pipeline/vimc/vimc.cpp
> @@ -114,6 +114,9 @@ static const std::map<PixelFormat, uint32_t> pixelformats{
>         { formats::BGR888, MEDIA_BUS_FMT_RGB888_1X24 },
>  };
>  
> +static constexpr Size kMinSize{ 16, 16 };
> +static constexpr Size kMaxSize{ 4096, 2160 };
> +
>  } /* namespace */
>  
>  VimcCameraConfiguration::VimcCameraConfiguration(VimcCameraData *data)
> @@ -153,14 +156,20 @@ CameraConfiguration::Status VimcCameraConfiguration::validate()
>         const Size size = cfg.size;
>  
>         /*
> -        * The scaler hardcodes a x3 scale-up ratio, and the sensor output size
> -        * is aligned to two pixels in both directions. The output width and
> -        * height thus have to be multiples of 6.
> +        * The sensor output size is aligned to two pixels in both directions.
> +        * Additionally, prior to v5.16, the scaler hardcodes a x3 scale-up
> +        * ratio, requiring the output width and height to be multiples of 6.
>          */
> -       cfg.size.width = std::max(48U, std::min(4096U, cfg.size.width));
> -       cfg.size.height = std::max(48U, std::min(2160U, cfg.size.height));
> -       cfg.size.width -= cfg.size.width % 6;
> -       cfg.size.height -= cfg.size.height % 6;
> +       Size minSize{ kMinSize };
> +       unsigned int alignment = 2;
> +
> +       if (data_->media_->version() < KERNEL_VERSION(5, 16, 0)) {
> +               minSize *= 3;
> +               alignment *= 3;
> +       }
> +
> +       cfg.size.expandTo(minSize).boundTo(kMaxSize)
> +               .alignDownTo(alignment, alignment);
>  
>         if (cfg.size != size) {
>                 LOG(VIMC, Debug)
> @@ -216,10 +225,12 @@ PipelineHandlerVimc::generateConfiguration(Camera *camera,
>                         }
>                 }
>  
> -               /* The scaler hardcodes a x3 scale-up ratio. */
> -               std::vector<SizeRange> sizes{
> -                       SizeRange{ { 48, 48 }, { 4096, 2160 } }
> -               };
> +               /* Prior to v5.16, the scaler hardcodes a x3 scale-up ratio. */
> +               Size minSize{ kMinSize };
> +               if (data->media_->version() < KERNEL_VERSION(5, 16, 0))
> +                       minSize *= 3;
> +
> +               std::vector<SizeRange> sizes{ { minSize, kMaxSize } };
>                 formats[pixelformat.first] = sizes;
>         }
>  
> @@ -242,10 +253,18 @@ int PipelineHandlerVimc::configure(Camera *camera, CameraConfiguration *config)
>         StreamConfiguration &cfg = config->at(0);
>         int ret;
>  
> -       /* The scaler hardcodes a x3 scale-up ratio. */
> +       /*
> +        * Prior to v5.16, the scaler hardcodes a x3 scale-up ratio. For newer
> +        * kernels, use a sensor resolution of 1920x1080 and let the scaler
> +        * produce the requested stream size.
> +        */
> +       Size sensorSize{ 1920, 1080 };
> +       if (data->media_->version() < KERNEL_VERSION(5, 16, 0))
> +               sensorSize = { cfg.size.width / 3, cfg.size.height / 3 };
> +
>         V4L2SubdeviceFormat subformat = {};
>         subformat.code = MEDIA_BUS_FMT_SGRBG8_1X8;
> -       subformat.size = { cfg.size.width / 3, cfg.size.height / 3 };
> +       subformat.size = sensorSize;
>  
>         ret = data->sensor_->setFormat(&subformat);
>         if (ret)
> @@ -293,7 +312,7 @@ int PipelineHandlerVimc::configure(Camera *camera, CameraConfiguration *config)
>          * vimc driver will fail pipeline validation.
>          */
>         format.fourcc = V4L2PixelFormat(V4L2_PIX_FMT_SGRBG8);
> -       format.size = { cfg.size.width / 3, cfg.size.height / 3 };
> +       format.size = sensorSize;
>  

Sounds pretty good to me. I haven't tested, but I expect you have.


Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>

>         ret = data->raw_->setFormat(&format);
>         if (ret)
> -- 
> Regards,
> 
> Laurent Pinchart
>