From: Randy Li <ayaka@xxxxxxxxxxx> Signed-off-by: Randy Li <ayaka@xxxxxxxxxxx> Signed-off-by: Hsia-Jun(Randy) Li <randy.li@xxxxxxxxxxxxx> --- Documentation/gpu/synaptics.rst | 81 +++++++++++++++++++++++++++++++++ 1 file changed, 81 insertions(+) create mode 100644 Documentation/gpu/synaptics.rst diff --git a/Documentation/gpu/synaptics.rst b/Documentation/gpu/synaptics.rst new file mode 100644 index 000000000000..4185ca536bf1 --- /dev/null +++ b/Documentation/gpu/synaptics.rst @@ -0,0 +1,81 @@ +.. SPDX-License-Identifier: GFDL-1.1-no-invariants-or-later + +================ +Synaptics Tiling +================ + +The tiling pixel formats in Synpatics Video Smart platform have +many variants. Tiles could form the group of tiles, pixels within +the group (nearest) width and height are stored into tile. +Meanwhile, the tile in a group may not follow dimension layout, +tile could form a small group of tiles, then that (sub)group +of tiles would form a bigger group. We won't describe the dimension +layout inside the group of tiles here. The layout of the group +of tiles is fixed with the group width and height parameters +in the same generation of the platform. + +Compression +=========== +The proprietary lossless image compression protocol in Synaptics +could minimizes the amount of data transferred (less memory bandwidth +consumption) between devices. It would usually apply to the tiling +pixel format. + +Each component would request an extra page aligned length buffer +for storing the compression meta data. Also a 32 bytes parameters +set would come with a compression meta data buffer. + +The component here corresponds to a signal type (i.e. Luma, chroma). +They could be encoded into one or multiple metadata planes, but +their compression parameters still would be individual. + +Pixel format modifiers +====================== +Addition alignment requirement for stride and size of a memory plane +could apply beyond what has been mentioned below. Remember always +negotiating with all the devices in pipeline before allocation. + +.. flat-table:: Synpatics Image Format Modifiers + + * - Identifier + - Fourcc + - Details + + * - DRM_FORMAT_MOD_SYNA_V4H1 + - DRM_FORMAT_NV12 + - The plain uncompressed 8 bits tile format. It sounds similar to + Intel's Y-tile. but it won't take any pixel from the next X direction + in a tile group. The line stride and image height must be aligned to + a multiple of 16. The height of chrominance plane would plus 8. + + * - DRM_FORMAT_MOD_SYNA_V4H3P8 + - DRM_FORMAT_NV15 + - The plain uncompressed 10 bits tile format. It stores pixel in 2D + 3x4 tiles with a 8bits padding to each of tile. Then a tile is in a + 128 bits cache line. + + * - DRM_FORMAT_MOD_SYNA_V4H1_64L4_COMPRESSED + - DRM_FORMAT_NV12 + - Group of tiles and compressed variant of ``DRM_FORMAT_MOD_SYNA_V4H1``. + A group of tiles would contain 64x4 pixels, where a tile has 1x4 + pixel. + + * - DRM_FORMAT_MOD_SYNA_V4H3P8_64L4_COMPRESSED + - DRM_FORMAT_NV15 + - Group of tiles and compressed variant of ``DRM_FORMAT_MOD_SYNA_V4H3P8``. + A group of tiles would contains 48x4 pixels, where a tile has 3x4 pixels + and a 8 bits padding in the end of a tile. A group of tiles would + be 256 bytes. + + * - ``DRM_FORMAT_MOD_SYNA_V4H1_128L128_COMPRESSED`` + - DRM_FORMAT_NV12 + - Group of tiles and compressed variant of ``DRM_FORMAT_MOD_SYNA_V4H1``. + A group of tiles would contain 128x32 pixels, where a tile has 1x4 + pixel. + + * - ``DRM_FORMAT_MOD_SYNA_V4H3P8_128L128_COMPRESSED`` + - DRM_FORMAT_NV15 + - Group of tiles and compressed variant of ``DRM_FORMAT_MOD_SYNA_V4H3P8``. + A group of tiles would contains 96x128 pixels, where a tile has 3x4 pixels + and a 8 bits padding in the end of a tile. A group of tiles would + be 16 KiB. -- 2.17.1
