On Mon, 27 Jun 2022 13:40:04 +0000 Dennis Tsiang <Dennis.Tsiang@xxxxxxx> wrote: > This patch is an early RFC to discuss the viable options and > alternatives for inclusion of unsigned integer formats for the DRM API. > > This patch adds a new single component 16-bit and a two component 32-bit > DRM fourcc’s that represent unsigned integer formats. The use case for > needing UINT formats, in our case, would be to support using raw buffers > for camera ISPs. > > For images imported with DRM fourcc + modifier combination, the GPU > driver needs a way to determine the datatype of the format which > currently the DRM API does not provide explicitly with a notable > exception of the floating-point fourccs such as DRM_FORMAT_XRGB16161616F > as an example. As the DRM fourccs do not currently define the > interpretation of the data, should the information be made explicit in > the DRM API similarly to how it is already done in Vulkan? > > The reason for introducing datatype to the DRM fourcc's is that the > alternative, for any API (e.g., EGL) that lacks the format datatype > information for fourcc/modifier combination for dma_buf interop would be > to introduce explicit additional metadata/attributes that encode this > information which then would be passed to the GPU driver but the > drawback of this is that it would require extending multiple graphics > APIs to support every single platform. > > By having the DRM API expose the datatype information for formats saves > a lot of integration/verification work for all of the different graphics > APIs and platforms as this information could be determined by the DRM > triple alone for dma_buf interop. > > It would be good to gather some opinions on what others think about > introducing datatypes to the DRM API. Hi, I didn't quite grasp where this information is necessary, and when it is necessary, is it that simple to communicate? Does it even belong with the pixel format at all? Let us consider the existing problems. All traditional integer formats in drm_fourcc.h right now are unsigned. They get interpreted as being in the range [0.0, 1.0] for color operations, which means converting to another bit depth works implicitly. That's where the simplicity ends. We assume to have full quantization range unless otherwise noted in some auxiliary data, like KMS properties (I forget if there even was a property to say DRM framebuffer uses limited quantization range). We assume all pixel data is non-linearly encoded. There is no color space information. YUV-RGB conversion matrix coefficients are handled by a KMS property IIRC. Coming back to the nominal range [0.0, 1.0]. That's an implicit assumption that allows us to apply things like LUTs. It already breaks down if you choose a floating-point format instead of unsigned integer format. Is FP pixel value 1.0 the same as nominal 1.0? Or is the FP pixel value 255.0 the same as nominal 1.0? KMS has no way to know or control that AFAIK. If you had UINT format, meaning the nominal value range is [0.0, 65535.0] (e.g. for 16 bpc) instead of [0.0, 1.0], then how does that work with a LUT element in the color pipeline? How would it be both meaningful and different to existing 16 bpc integer format? What's the actual difference between R16 and R16_UINT, what difference does it make to a GPU driver? Maybe that is intimately dependent on the API where the pixel formats are used? Maybe for KMS R16 and R16_UINT would be completely equivalent, but not for some other API? We also need to be very careful to not load pixel format with meaning that does not belong there. Non-linear encoding (transfer function) is obviously something that is completely unrelated to the pixel format, as long as the pixel format defines a conversion to nominal value range. Color space (primaries and white point) are another thing that has nothing to do with pixel format, and so must not be in any way implied by pixel format. Should a pixel format define how the raw pixel values are converted to nominal values? No, because we have quantization range as a separate property, currently with "full" and "limited" understood, where "limited" means different things depending on the color model. Color model is defined by the pixel formats: we have RGB and YUV formats, likewise is chroma sub-sampling. Hmm. Thanks, pq
Attachment:
pgpPN7Aa20_0W.pgp
Description: OpenPGP digital signature
