[libcamera-devel] [PATCH 02/11] include: linux: Add ipu3 kernel header and format definitions
Jacopo Mondi
jacopo at jmondi.org
Mon Nov 9 10:57:51 CET 2020
Hi Niklas,
On Thu, Nov 05, 2020 at 01:15:37AM +0100, Niklas Söderlund wrote:
> Add kernel header and format definitions for the Intel IPU3. The headers
> have been exported from a v5.9 kernel using scripts/headers_install.sh.
>
> Signed-off-by: Niklas Söderlund <niklas.soderlund at ragnatech.se>
If, as I got it, the header comes form the v5.9 tag there's no need
for a SHA1, otherwise it might be opportune to add one.
Anyway
Reviewed-by: Jacopo Mondi <jacopo at jmondi.org>
Thanks
j
> ---
> include/linux/intel-ipu3.h | 2785 ++++++++++++++++++++++++++++++++++++
> 1 file changed, 2785 insertions(+)
> create mode 100644 include/linux/intel-ipu3.h
>
> diff --git a/include/linux/intel-ipu3.h b/include/linux/intel-ipu3.h
> new file mode 100644
> index 0000000000000000..8ed1632164fddf44
> --- /dev/null
> +++ b/include/linux/intel-ipu3.h
> @@ -0,0 +1,2785 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/* Copyright (C) 2017 - 2018 Intel Corporation */
> +
> +#ifndef __IPU3_UAPI_H
> +#define __IPU3_UAPI_H
> +
> +#include <linux/types.h>
> +
> +/* from /drivers/staging/media/ipu3/include/videodev2.h */
> +
> +/* Vendor specific - used for IPU3 camera sub-system */
> +#define V4L2_META_FMT_IPU3_PARAMS v4l2_fourcc('i', 'p', '3', 'p') /* IPU3 processing parameters */
> +#define V4L2_META_FMT_IPU3_STAT_3A v4l2_fourcc('i', 'p', '3', 's') /* IPU3 3A statistics */
> +
> +/* from include/uapi/linux/v4l2-controls.h */
> +#define V4L2_CID_INTEL_IPU3_BASE (V4L2_CID_USER_BASE + 0x10c0)
> +#define V4L2_CID_INTEL_IPU3_MODE (V4L2_CID_INTEL_IPU3_BASE + 1)
> +
> +/******************* ipu3_uapi_stats_3a *******************/
> +
> +#define IPU3_UAPI_MAX_STRIPES 2
> +#define IPU3_UAPI_MAX_BUBBLE_SIZE 10
> +
> +#define IPU3_UAPI_GRID_START_MASK ((1 << 12) - 1)
> +#define IPU3_UAPI_GRID_Y_START_EN (1 << 15)
> +
> +/* controls generation of meta_data (like FF enable/disable) */
> +#define IPU3_UAPI_AWB_RGBS_THR_B_EN (1 << 14)
> +#define IPU3_UAPI_AWB_RGBS_THR_B_INCL_SAT (1 << 15)
> +
> +/**
> + * struct ipu3_uapi_grid_config - Grid plane config
> + *
> + * @width: Grid horizontal dimensions, in number of grid blocks(cells).
> + * @height: Grid vertical dimensions, in number of grid cells.
> + * @block_width_log2: Log2 of the width of each cell in pixels.
> + * for (2^3, 2^4, 2^5, 2^6, 2^7), values [3, 7].
> + * @block_height_log2: Log2 of the height of each cell in pixels.
> + * for (2^3, 2^4, 2^5, 2^6, 2^7), values [3, 7].
> + * @height_per_slice: The number of blocks in vertical axis per slice.
> + * Default 2.
> + * @x_start: X value of top left corner of Region of Interest(ROI).
> + * @y_start: Y value of top left corner of ROI
> + * @x_end: X value of bottom right corner of ROI
> + * @y_end: Y value of bottom right corner of ROI
> + *
> + * Due to the size of total amount of collected data, most statistics
> + * create a grid-based output, and the data is then divided into "slices".
> + */
> +struct ipu3_uapi_grid_config {
> + __u8 width;
> + __u8 height;
> + __u16 block_width_log2:3;
> + __u16 block_height_log2:3;
> + __u16 height_per_slice:8;
> + __u16 x_start;
> + __u16 y_start;
> + __u16 x_end;
> + __u16 y_end;
> +} __attribute__((packed));
> +
> +/*
> + * The grid based data is divided into "slices" called set, each slice of setX
> + * refers to ipu3_uapi_grid_config width * height_per_slice.
> + */
> +#define IPU3_UAPI_AWB_MAX_SETS 60
> +/* Based on grid size 80 * 60 and cell size 16 x 16 */
> +#define IPU3_UAPI_AWB_SET_SIZE 1280
> +#define IPU3_UAPI_AWB_MD_ITEM_SIZE 8
> +#define IPU3_UAPI_AWB_SPARE_FOR_BUBBLES \
> + (IPU3_UAPI_MAX_BUBBLE_SIZE * IPU3_UAPI_MAX_STRIPES * \
> + IPU3_UAPI_AWB_MD_ITEM_SIZE)
> +#define IPU3_UAPI_AWB_MAX_BUFFER_SIZE \
> + (IPU3_UAPI_AWB_MAX_SETS * \
> + (IPU3_UAPI_AWB_SET_SIZE + IPU3_UAPI_AWB_SPARE_FOR_BUBBLES))
> +
> +
> +/**
> + * struct ipu3_uapi_awb_raw_buffer - AWB raw buffer
> + *
> + * @meta_data: buffer to hold auto white balance meta data which is
> + * the average values for each color channel.
> + */
> +struct ipu3_uapi_awb_raw_buffer {
> + __u8 meta_data[IPU3_UAPI_AWB_MAX_BUFFER_SIZE]
> + __attribute__((aligned(32)));
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_awb_config_s - AWB config
> + *
> + * @rgbs_thr_gr: gr threshold value.
> + * @rgbs_thr_r: Red threshold value.
> + * @rgbs_thr_gb: gb threshold value.
> + * @rgbs_thr_b: Blue threshold value.
> + * @grid: &ipu3_uapi_grid_config, the default grid resolution is 16x16 cells.
> + *
> + * The threshold is a saturation measure range [0, 8191], 8191 is default.
> + * Values over threshold may be optionally rejected for averaging.
> + */
> +struct ipu3_uapi_awb_config_s {
> + __u16 rgbs_thr_gr;
> + __u16 rgbs_thr_r;
> + __u16 rgbs_thr_gb;
> + __u16 rgbs_thr_b;
> + struct ipu3_uapi_grid_config grid;
> +} __attribute__((aligned(32))) __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_awb_config - AWB config wrapper
> + *
> + * @config: config for auto white balance as defined by &ipu3_uapi_awb_config_s
> + */
> +struct ipu3_uapi_awb_config {
> + struct ipu3_uapi_awb_config_s config __attribute__((aligned(32)));
> +} __attribute__((packed));
> +
> +#define IPU3_UAPI_AE_COLORS 4 /* R, G, B, Y */
> +#define IPU3_UAPI_AE_BINS 256
> +#define IPU3_UAPI_AE_WEIGHTS 96
> +
> +/**
> + + * struct ipu3_uapi_ae_raw_buffer - AE global weighted histogram
> + + *
> + + * @vals: Sum of IPU3_UAPI_AE_COLORS in cell
> + + *
> + + * Each histogram contains IPU3_UAPI_AE_BINS bins. Each bin has 24 bit unsigned
> + + * for counting the number of the pixel.
> + + */
> +struct ipu3_uapi_ae_raw_buffer {
> + __u32 vals[IPU3_UAPI_AE_BINS * IPU3_UAPI_AE_COLORS];
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_ae_raw_buffer_aligned - AE raw buffer
> + *
> + * @buff: &ipu3_uapi_ae_raw_buffer to hold full frame meta data.
> + */
> +struct ipu3_uapi_ae_raw_buffer_aligned {
> + struct ipu3_uapi_ae_raw_buffer buff __attribute__((aligned(32)));
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_ae_grid_config - AE weight grid
> + *
> + * @width: Grid horizontal dimensions. Value: [16, 32], default 16.
> + * @height: Grid vertical dimensions. Value: [16, 24], default 16.
> + * @block_width_log2: Log2 of the width of the grid cell, value: [3, 7].
> + * @block_height_log2: Log2 of the height of the grid cell, value: [3, 7].
> + * default is 3 (cell size 8x8), 4 cell per grid.
> + * @reserved0: reserved
> + * @ae_en: 0: does not write to &ipu3_uapi_ae_raw_buffer_aligned array,
> + * 1: write normally.
> + * @rst_hist_array: write 1 to trigger histogram array reset.
> + * @done_rst_hist_array: flag for histogram array reset done.
> + * @x_start: X value of top left corner of ROI, default 0.
> + * @y_start: Y value of top left corner of ROI, default 0.
> + * @x_end: X value of bottom right corner of ROI
> + * @y_end: Y value of bottom right corner of ROI
> + *
> + * The AE block accumulates 4 global weighted histograms(R, G, B, Y) over
> + * a defined ROI within the frame. The contribution of each pixel into the
> + * histogram, defined by &ipu3_uapi_ae_weight_elem LUT, is indexed by a grid.
> + */
> +struct ipu3_uapi_ae_grid_config {
> + __u8 width;
> + __u8 height;
> + __u8 block_width_log2:4;
> + __u8 block_height_log2:4;
> + __u8 reserved0:5;
> + __u8 ae_en:1;
> + __u8 rst_hist_array:1;
> + __u8 done_rst_hist_array:1;
> + __u16 x_start;
> + __u16 y_start;
> + __u16 x_end;
> + __u16 y_end;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_ae_weight_elem - AE weights LUT
> + *
> + * @cell0: weighted histogram grid value.
> + * @cell1: weighted histogram grid value.
> + * @cell2: weighted histogram grid value.
> + * @cell3: weighted histogram grid value.
> + * @cell4: weighted histogram grid value.
> + * @cell5: weighted histogram grid value.
> + * @cell6: weighted histogram grid value.
> + * @cell7: weighted histogram grid value.
> + *
> + * Use weighted grid value to give a different contribution factor to each cell.
> + * Precision u4, range [0, 15].
> + */
> +struct ipu3_uapi_ae_weight_elem {
> + __u32 cell0:4;
> + __u32 cell1:4;
> + __u32 cell2:4;
> + __u32 cell3:4;
> + __u32 cell4:4;
> + __u32 cell5:4;
> + __u32 cell6:4;
> + __u32 cell7:4;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_ae_ccm - AE coefficients for WB and CCM
> + *
> + * @gain_gr: WB gain factor for the gr channels. Default 256.
> + * @gain_r: WB gain factor for the r channel. Default 256.
> + * @gain_b: WB gain factor for the b channel. Default 256.
> + * @gain_gb: WB gain factor for the gb channels. Default 256.
> + * @mat: 4x4 matrix that transforms Bayer quad output from WB to RGB+Y.
> + *
> + * Default:
> + * 128, 0, 0, 0,
> + * 0, 128, 0, 0,
> + * 0, 0, 128, 0,
> + * 0, 0, 0, 128,
> + *
> + * As part of the raw frame pre-process stage, the WB and color conversion need
> + * to be applied to expose the impact of these gain operations.
> + */
> +struct ipu3_uapi_ae_ccm {
> + __u16 gain_gr;
> + __u16 gain_r;
> + __u16 gain_b;
> + __u16 gain_gb;
> + __s16 mat[16];
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_ae_config - AE config
> + *
> + * @grid_cfg: config for auto exposure statistics grid. See struct
> + * &ipu3_uapi_ae_grid_config
> + * @weights: &IPU3_UAPI_AE_WEIGHTS is based on 32x24 blocks in the grid.
> + * Each grid cell has a corresponding value in weights LUT called
> + * grid value, global histogram is updated based on grid value and
> + * pixel value.
> + * @ae_ccm: Color convert matrix pre-processing block.
> + *
> + * Calculate AE grid from image resolution, resample ae weights.
> + */
> +struct ipu3_uapi_ae_config {
> + struct ipu3_uapi_ae_grid_config grid_cfg __attribute__((aligned(32)));
> + struct ipu3_uapi_ae_weight_elem weights[
> + IPU3_UAPI_AE_WEIGHTS] __attribute__((aligned(32)));
> + struct ipu3_uapi_ae_ccm ae_ccm __attribute__((aligned(32)));
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_af_filter_config - AF 2D filter for contrast measurements
> + *
> + * @y1_coeff_0: filter Y1, structure: 3x11, support both symmetry and
> + * anti-symmetry type. A12 is center, A1-A11 are neighbours.
> + * for analyzing low frequency content, used to calculate sum
> + * of gradients in x direction.
> + * @y1_coeff_0.a1: filter1 coefficients A1, u8, default 0.
> + * @y1_coeff_0.a2: filter1 coefficients A2, u8, default 0.
> + * @y1_coeff_0.a3: filter1 coefficients A3, u8, default 0.
> + * @y1_coeff_0.a4: filter1 coefficients A4, u8, default 0.
> + * @y1_coeff_1: Struct
> + * @y1_coeff_1.a5: filter1 coefficients A5, u8, default 0.
> + * @y1_coeff_1.a6: filter1 coefficients A6, u8, default 0.
> + * @y1_coeff_1.a7: filter1 coefficients A7, u8, default 0.
> + * @y1_coeff_1.a8: filter1 coefficients A8, u8, default 0.
> + * @y1_coeff_2: Struct
> + * @y1_coeff_2.a9: filter1 coefficients A9, u8, default 0.
> + * @y1_coeff_2.a10: filter1 coefficients A10, u8, default 0.
> + * @y1_coeff_2.a11: filter1 coefficients A11, u8, default 0.
> + * @y1_coeff_2.a12: filter1 coefficients A12, u8, default 128.
> + * @y1_sign_vec: Each bit corresponds to one coefficient sign bit,
> + * 0: positive, 1: negative, default 0.
> + * @y2_coeff_0: Y2, same structure as Y1. For analyzing high frequency content.
> + * @y2_coeff_0.a1: filter2 coefficients A1, u8, default 0.
> + * @y2_coeff_0.a2: filter2 coefficients A2, u8, default 0.
> + * @y2_coeff_0.a3: filter2 coefficients A3, u8, default 0.
> + * @y2_coeff_0.a4: filter2 coefficients A4, u8, default 0.
> + * @y2_coeff_1: Struct
> + * @y2_coeff_1.a5: filter2 coefficients A5, u8, default 0.
> + * @y2_coeff_1.a6: filter2 coefficients A6, u8, default 0.
> + * @y2_coeff_1.a7: filter2 coefficients A7, u8, default 0.
> + * @y2_coeff_1.a8: filter2 coefficients A8, u8, default 0.
> + * @y2_coeff_2: Struct
> + * @y2_coeff_2.a9: filter1 coefficients A9, u8, default 0.
> + * @y2_coeff_2.a10: filter1 coefficients A10, u8, default 0.
> + * @y2_coeff_2.a11: filter1 coefficients A11, u8, default 0.
> + * @y2_coeff_2.a12: filter1 coefficients A12, u8, default 128.
> + * @y2_sign_vec: Each bit corresponds to one coefficient sign bit,
> + * 0: positive, 1: negative, default 0.
> + * @y_calc: Pre-processing that converts Bayer quad to RGB+Y values to be
> + * used for building histogram. Range [0, 32], default 8.
> + * Rule:
> + * y_gen_rate_gr + y_gen_rate_r + y_gen_rate_b + y_gen_rate_gb = 32
> + * A single Y is calculated based on sum of Gr/R/B/Gb based on
> + * their contribution ratio.
> + * @y_calc.y_gen_rate_gr: Contribution ratio Gr for Y
> + * @y_calc.y_gen_rate_r: Contribution ratio R for Y
> + * @y_calc.y_gen_rate_b: Contribution ratio B for Y
> + * @y_calc.y_gen_rate_gb: Contribution ratio Gb for Y
> + * @nf: The shift right value that should be applied during the Y1/Y2 filter to
> + * make sure the total memory needed is 2 bytes per grid cell.
> + * @nf.reserved0: reserved
> + * @nf.y1_nf: Normalization factor for the convolution coeffs of y1,
> + * should be log2 of the sum of the abs values of the filter
> + * coeffs, default 7 (2^7 = 128).
> + * @nf.reserved1: reserved
> + * @nf.y2_nf: Normalization factor for y2, should be log2 of the sum of the
> + * abs values of the filter coeffs.
> + * @nf.reserved2: reserved
> + */
> +struct ipu3_uapi_af_filter_config {
> + struct {
> + __u8 a1;
> + __u8 a2;
> + __u8 a3;
> + __u8 a4;
> + } y1_coeff_0;
> + struct {
> + __u8 a5;
> + __u8 a6;
> + __u8 a7;
> + __u8 a8;
> + } y1_coeff_1;
> + struct {
> + __u8 a9;
> + __u8 a10;
> + __u8 a11;
> + __u8 a12;
> + } y1_coeff_2;
> +
> + __u32 y1_sign_vec;
> +
> + struct {
> + __u8 a1;
> + __u8 a2;
> + __u8 a3;
> + __u8 a4;
> + } y2_coeff_0;
> + struct {
> + __u8 a5;
> + __u8 a6;
> + __u8 a7;
> + __u8 a8;
> + } y2_coeff_1;
> + struct {
> + __u8 a9;
> + __u8 a10;
> + __u8 a11;
> + __u8 a12;
> + } y2_coeff_2;
> +
> + __u32 y2_sign_vec;
> +
> + struct {
> + __u8 y_gen_rate_gr;
> + __u8 y_gen_rate_r;
> + __u8 y_gen_rate_b;
> + __u8 y_gen_rate_gb;
> + } y_calc;
> +
> + struct {
> + __u32 reserved0:8;
> + __u32 y1_nf:4;
> + __u32 reserved1:4;
> + __u32 y2_nf:4;
> + __u32 reserved2:12;
> + } nf;
> +} __attribute__((packed));
> +
> +#define IPU3_UAPI_AF_MAX_SETS 24
> +#define IPU3_UAPI_AF_MD_ITEM_SIZE 4
> +#define IPU3_UAPI_AF_SPARE_FOR_BUBBLES \
> + (IPU3_UAPI_MAX_BUBBLE_SIZE * IPU3_UAPI_MAX_STRIPES * \
> + IPU3_UAPI_AF_MD_ITEM_SIZE)
> +#define IPU3_UAPI_AF_Y_TABLE_SET_SIZE 128
> +#define IPU3_UAPI_AF_Y_TABLE_MAX_SIZE \
> + (IPU3_UAPI_AF_MAX_SETS * \
> + (IPU3_UAPI_AF_Y_TABLE_SET_SIZE + IPU3_UAPI_AF_SPARE_FOR_BUBBLES) * \
> + IPU3_UAPI_MAX_STRIPES)
> +
> +/**
> + * struct ipu3_uapi_af_raw_buffer - AF meta data
> + *
> + * @y_table: Each color component will be convolved separately with filter1
> + * and filter2 and the result will be summed out and averaged for
> + * each cell.
> + */
> +struct ipu3_uapi_af_raw_buffer {
> + __u8 y_table[IPU3_UAPI_AF_Y_TABLE_MAX_SIZE] __attribute__((aligned(32)));
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_af_config_s - AF config
> + *
> + * @filter_config: AF uses Y1 and Y2 filters as configured in
> + * &ipu3_uapi_af_filter_config
> + * @padding: paddings
> + * @grid_cfg: See &ipu3_uapi_grid_config, default resolution 16x16. Use large
> + * grid size for large image and vice versa.
> + */
> +struct ipu3_uapi_af_config_s {
> + struct ipu3_uapi_af_filter_config filter_config __attribute__((aligned(32)));
> + __u8 padding[4];
> + struct ipu3_uapi_grid_config grid_cfg __attribute__((aligned(32)));
> +} __attribute__((packed));
> +
> +#define IPU3_UAPI_AWB_FR_MAX_SETS 24
> +#define IPU3_UAPI_AWB_FR_MD_ITEM_SIZE 8
> +#define IPU3_UAPI_AWB_FR_BAYER_TBL_SIZE 256
> +#define IPU3_UAPI_AWB_FR_SPARE_FOR_BUBBLES \
> + (IPU3_UAPI_MAX_BUBBLE_SIZE * IPU3_UAPI_MAX_STRIPES * \
> + IPU3_UAPI_AWB_FR_MD_ITEM_SIZE)
> +#define IPU3_UAPI_AWB_FR_BAYER_TABLE_MAX_SIZE \
> + (IPU3_UAPI_AWB_FR_MAX_SETS * \
> + (IPU3_UAPI_AWB_FR_BAYER_TBL_SIZE + \
> + IPU3_UAPI_AWB_FR_SPARE_FOR_BUBBLES) * IPU3_UAPI_MAX_STRIPES)
> +
> +/**
> + * struct ipu3_uapi_awb_fr_meta_data - AWB filter response meta data
> + *
> + * @meta_data: Statistics output on the grid after convolving with 1D filter.
> + */
> +struct ipu3_uapi_awb_fr_raw_buffer {
> + __u8 meta_data[IPU3_UAPI_AWB_FR_BAYER_TABLE_MAX_SIZE]
> + __attribute__((aligned(32)));
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_awb_fr_config_s - AWB filter response config
> + *
> + * @grid_cfg: grid config, default 16x16.
> + * @bayer_coeff: 1D Filter 1x11 center symmetry/anti-symmetry.
> + * coefficients defaults { 0, 0, 0, 0, 0, 128 }.
> + * Applied on whole image for each Bayer channel separately
> + * by a weighted sum of its 11x1 neighbors.
> + * @reserved1: reserved
> + * @bayer_sign: sign of filter coefficients, default 0.
> + * @bayer_nf: normalization factor for the convolution coeffs, to make sure
> + * total memory needed is within pre-determined range.
> + * NF should be the log2 of the sum of the abs values of the
> + * filter coeffs, range [7, 14], default 7.
> + * @reserved2: reserved
> + */
> +struct ipu3_uapi_awb_fr_config_s {
> + struct ipu3_uapi_grid_config grid_cfg;
> + __u8 bayer_coeff[6];
> + __u16 reserved1;
> + __u32 bayer_sign;
> + __u8 bayer_nf;
> + __u8 reserved2[7];
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_4a_config - 4A config
> + *
> + * @awb_config: &ipu3_uapi_awb_config_s, default resolution 16x16
> + * @ae_grd_config: auto exposure statistics &ipu3_uapi_ae_grid_config
> + * @padding: paddings
> + * @af_config: auto focus config &ipu3_uapi_af_config_s
> + * @awb_fr_config: &ipu3_uapi_awb_fr_config_s, default resolution 16x16
> + */
> +struct ipu3_uapi_4a_config {
> + struct ipu3_uapi_awb_config_s awb_config __attribute__((aligned(32)));
> + struct ipu3_uapi_ae_grid_config ae_grd_config;
> + __u8 padding[20];
> + struct ipu3_uapi_af_config_s af_config;
> + struct ipu3_uapi_awb_fr_config_s awb_fr_config
> + __attribute__((aligned(32)));
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_bubble_info - Bubble info for host side debugging
> + *
> + * @num_of_stripes: A single frame is divided into several parts called stripes
> + * due to limitation on line buffer memory.
> + * The separation between the stripes is vertical. Each such
> + * stripe is processed as a single frame by the ISP pipe.
> + * @padding: padding bytes.
> + * @num_sets: number of sets.
> + * @padding1: padding bytes.
> + * @size_of_set: set size.
> + * @padding2: padding bytes.
> + * @bubble_size: is the amount of padding in the bubble expressed in "sets".
> + * @padding3: padding bytes.
> + */
> +struct ipu3_uapi_bubble_info {
> + __u32 num_of_stripes __attribute__((aligned(32)));
> + __u8 padding[28];
> + __u32 num_sets;
> + __u8 padding1[28];
> + __u32 size_of_set;
> + __u8 padding2[28];
> + __u32 bubble_size;
> + __u8 padding3[28];
> +} __attribute__((packed));
> +
> +/*
> + * struct ipu3_uapi_stats_3a_bubble_info_per_stripe
> + */
> +struct ipu3_uapi_stats_3a_bubble_info_per_stripe {
> + struct ipu3_uapi_bubble_info awb[IPU3_UAPI_MAX_STRIPES];
> + struct ipu3_uapi_bubble_info af[IPU3_UAPI_MAX_STRIPES];
> + struct ipu3_uapi_bubble_info awb_fr[IPU3_UAPI_MAX_STRIPES];
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_ff_status - Enable bits for each 3A fixed function
> + *
> + * @awb_en: auto white balance enable
> + * @padding: padding config
> + * @ae_en: auto exposure enable
> + * @padding1: padding config
> + * @af_en: auto focus enable
> + * @padding2: padding config
> + * @awb_fr_en: awb filter response enable bit
> + * @padding3: padding config
> + */
> +struct ipu3_uapi_ff_status {
> + __u32 awb_en __attribute__((aligned(32)));
> + __u8 padding[28];
> + __u32 ae_en;
> + __u8 padding1[28];
> + __u32 af_en;
> + __u8 padding2[28];
> + __u32 awb_fr_en;
> + __u8 padding3[28];
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_stats_3a - 3A statistics
> + *
> + * @awb_raw_buffer: auto white balance meta data &ipu3_uapi_awb_raw_buffer
> + * @ae_raw_buffer: auto exposure raw data &ipu3_uapi_ae_raw_buffer_aligned
> + * @af_raw_buffer: &ipu3_uapi_af_raw_buffer for auto focus meta data
> + * @awb_fr_raw_buffer: value as specified by &ipu3_uapi_awb_fr_raw_buffer
> + * @stats_4a_config: 4a statistics config as defined by &ipu3_uapi_4a_config.
> + * @ae_join_buffers: 1 to use ae_raw_buffer.
> + * @padding: padding config
> + * @stats_3a_bubble_per_stripe: a &ipu3_uapi_stats_3a_bubble_info_per_stripe
> + * @stats_3a_status: 3a statistics status set in &ipu3_uapi_ff_status
> + */
> +struct ipu3_uapi_stats_3a {
> + struct ipu3_uapi_awb_raw_buffer awb_raw_buffer;
> + struct ipu3_uapi_ae_raw_buffer_aligned
> + ae_raw_buffer[IPU3_UAPI_MAX_STRIPES];
> + struct ipu3_uapi_af_raw_buffer af_raw_buffer;
> + struct ipu3_uapi_awb_fr_raw_buffer awb_fr_raw_buffer;
> + struct ipu3_uapi_4a_config stats_4a_config;
> + __u32 ae_join_buffers;
> + __u8 padding[28];
> + struct ipu3_uapi_stats_3a_bubble_info_per_stripe
> + stats_3a_bubble_per_stripe;
> + struct ipu3_uapi_ff_status stats_3a_status;
> +} __attribute__((packed));
> +
> +/******************* ipu3_uapi_acc_param *******************/
> +
> +#define IPU3_UAPI_ISP_VEC_ELEMS 64
> +#define IPU3_UAPI_ISP_TNR3_VMEM_LEN 9
> +
> +#define IPU3_UAPI_BNR_LUT_SIZE 32
> +
> +/* number of elements in gamma correction LUT */
> +#define IPU3_UAPI_GAMMA_CORR_LUT_ENTRIES 256
> +
> +/* largest grid is 73x56, for grid_height_per_slice of 2, 73x2 = 146 */
> +#define IPU3_UAPI_SHD_MAX_CELLS_PER_SET 146
> +#define IPU3_UAPI_SHD_MAX_CFG_SETS 28
> +/* Normalization shift aka nf */
> +#define IPU3_UAPI_SHD_BLGR_NF_SHIFT 13
> +#define IPU3_UAPI_SHD_BLGR_NF_MASK 7
> +
> +#define IPU3_UAPI_YUVP2_TCC_MACC_TABLE_ELEMENTS 16
> +#define IPU3_UAPI_YUVP2_TCC_INV_Y_LUT_ELEMENTS 14
> +#define IPU3_UAPI_YUVP2_TCC_GAIN_PCWL_LUT_ELEMENTS 258
> +#define IPU3_UAPI_YUVP2_TCC_R_SQR_LUT_ELEMENTS 24
> +
> +#define IPU3_UAPI_ANR_LUT_SIZE 26
> +#define IPU3_UAPI_ANR_PYRAMID_SIZE 22
> +
> +#define IPU3_UAPI_LIN_LUT_SIZE 64
> +
> +/* Bayer Noise Reduction related structs */
> +
> +/**
> + * struct ipu3_uapi_bnr_static_config_wb_gains_config - White balance gains
> + *
> + * @gr: white balance gain for Gr channel.
> + * @r: white balance gain for R channel.
> + * @b: white balance gain for B channel.
> + * @gb: white balance gain for Gb channel.
> + *
> + * Precision u3.13, range [0, 8). White balance correction is done by applying
> + * a multiplicative gain to each color channels prior to BNR.
> + */
> +struct ipu3_uapi_bnr_static_config_wb_gains_config {
> + __u16 gr;
> + __u16 r;
> + __u16 b;
> + __u16 gb;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_bnr_static_config_wb_gains_thr_config - Threshold config
> + *
> + * @gr: white balance threshold gain for Gr channel.
> + * @r: white balance threshold gain for R channel.
> + * @b: white balance threshold gain for B channel.
> + * @gb: white balance threshold gain for Gb channel.
> + *
> + * Defines the threshold that specifies how different a defect pixel can be from
> + * its neighbors.(used by dynamic defect pixel correction sub block)
> + * Precision u4.4 range [0, 8].
> + */
> +struct ipu3_uapi_bnr_static_config_wb_gains_thr_config {
> + __u8 gr;
> + __u8 r;
> + __u8 b;
> + __u8 gb;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_bnr_static_config_thr_coeffs_config - Noise model
> + * coefficients that controls noise threshold
> + *
> + * @cf: Free coefficient for threshold calculation, range [0, 8191], default 0.
> + * @reserved0: reserved
> + * @cg: Gain coefficient for threshold calculation, [0, 31], default 8.
> + * @ci: Intensity coefficient for threshold calculation. range [0, 0x1f]
> + * default 6.
> + * format: u3.2 (3 most significant bits represent whole number,
> + * 2 least significant bits represent the fractional part
> + * with each count representing 0.25)
> + * e.g. 6 in binary format is 00110, that translates to 1.5
> + * @reserved1: reserved
> + * @r_nf: Normalization shift value for r^2 calculation, range [12, 20]
> + * where r is a radius of pixel [row, col] from centor of sensor.
> + * default 14.
> + *
> + * Threshold used to distinguish between noise and details.
> + */
> +struct ipu3_uapi_bnr_static_config_thr_coeffs_config {
> + __u32 cf:13;
> + __u32 reserved0:3;
> + __u32 cg:5;
> + __u32 ci:5;
> + __u32 reserved1:1;
> + __u32 r_nf:5;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_bnr_static_config_thr_ctrl_shd_config - Shading config
> + *
> + * @gr: Coefficient defines lens shading gain approximation for gr channel
> + * @r: Coefficient defines lens shading gain approximation for r channel
> + * @b: Coefficient defines lens shading gain approximation for b channel
> + * @gb: Coefficient defines lens shading gain approximation for gb channel
> + *
> + * Parameters for noise model (NM) adaptation of BNR due to shading correction.
> + * All above have precision of u3.3, default to 0.
> + */
> +struct ipu3_uapi_bnr_static_config_thr_ctrl_shd_config {
> + __u8 gr;
> + __u8 r;
> + __u8 b;
> + __u8 gb;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_bnr_static_config_opt_center_config - Optical center config
> + *
> + * @x_reset: Reset value of X (col start - X center). Precision s12.0.
> + * @reserved0: reserved
> + * @y_reset: Reset value of Y (row start - Y center). Precision s12.0.
> + * @reserved2: reserved
> + *
> + * Distance from corner to optical center for NM adaptation due to shading
> + * correction (should be calculated based on shading tables)
> + */
> +struct ipu3_uapi_bnr_static_config_opt_center_config {
> + __s32 x_reset:13;
> + __u32 reserved0:3;
> + __s32 y_reset:13;
> + __u32 reserved2:3;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_bnr_static_config_lut_config - BNR square root lookup table
> + *
> + * @values: pre-calculated values of square root function.
> + *
> + * LUT implementation of square root operation.
> + */
> +struct ipu3_uapi_bnr_static_config_lut_config {
> + __u8 values[IPU3_UAPI_BNR_LUT_SIZE];
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_bnr_static_config_bp_ctrl_config - Detect bad pixels (bp)
> + *
> + * @bp_thr_gain: Defines the threshold that specifies how different a
> + * defect pixel can be from its neighbors. Threshold is
> + * dependent on de-noise threshold calculated by algorithm.
> + * Range [4, 31], default 4.
> + * @reserved0: reserved
> + * @defect_mode: Mode of addressed defect pixels,
> + * 0 - single defect pixel is expected,
> + * 1 - 2 adjacent defect pixels are expected, default 1.
> + * @bp_gain: Defines how 2nd derivation that passes through a defect pixel
> + * is different from 2nd derivations that pass through
> + * neighbor pixels. u4.2, range [0, 256], default 8.
> + * @reserved1: reserved
> + * @w0_coeff: Blending coefficient of defect pixel correction.
> + * Precision u4, range [0, 8], default 8.
> + * @reserved2: reserved
> + * @w1_coeff: Enable influence of incorrect defect pixel correction to be
> + * avoided. Precision u4, range [1, 8], default 8.
> + * @reserved3: reserved
> + */
> +struct ipu3_uapi_bnr_static_config_bp_ctrl_config {
> + __u32 bp_thr_gain:5;
> + __u32 reserved0:2;
> + __u32 defect_mode:1;
> + __u32 bp_gain:6;
> + __u32 reserved1:18;
> + __u32 w0_coeff:4;
> + __u32 reserved2:4;
> + __u32 w1_coeff:4;
> + __u32 reserved3:20;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_bnr_static_config_dn_detect_ctrl_config - Denoising config
> + *
> + * @alpha: Weight of central element of smoothing filter.
> + * @beta: Weight of peripheral elements of smoothing filter, default 4.
> + * @gamma: Weight of diagonal elements of smoothing filter, default 4.
> + *
> + * beta and gamma parameter define the strength of the noise removal filter.
> + * All above has precision u0.4, range [0, 0xf]
> + * format: u0.4 (no / zero bits represent whole number,
> + * 4 bits represent the fractional part
> + * with each count representing 0.0625)
> + * e.g. 0xf translates to 0.0625x15 = 0.9375
> + *
> + * @reserved0: reserved
> + * @max_inf: Maximum increase of peripheral or diagonal element influence
> + * relative to the pre-defined value range: [0x5, 0xa]
> + * @reserved1: reserved
> + * @gd_enable: Green disparity enable control, 0 - disable, 1 - enable.
> + * @bpc_enable: Bad pixel correction enable control, 0 - disable, 1 - enable.
> + * @bnr_enable: Bayer noise removal enable control, 0 - disable, 1 - enable.
> + * @ff_enable: Fixed function enable, 0 - disable, 1 - enable.
> + * @reserved2: reserved
> + */
> +struct ipu3_uapi_bnr_static_config_dn_detect_ctrl_config {
> + __u32 alpha:4;
> + __u32 beta:4;
> + __u32 gamma:4;
> + __u32 reserved0:4;
> + __u32 max_inf:4;
> + __u32 reserved1:7;
> + __u32 gd_enable:1;
> + __u32 bpc_enable:1;
> + __u32 bnr_enable:1;
> + __u32 ff_enable:1;
> + __u32 reserved2:1;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_bnr_static_config_opt_center_sqr_config - BNR optical square
> + *
> + * @x_sqr_reset: Reset value of X^2.
> + * @y_sqr_reset: Reset value of Y^2.
> + *
> + * Please note:
> + *
> + * #. X and Y ref to
> + * &ipu3_uapi_bnr_static_config_opt_center_config
> + * #. Both structs are used in threshold formula to calculate r^2, where r
> + * is a radius of pixel [row, col] from centor of sensor.
> + */
> +struct ipu3_uapi_bnr_static_config_opt_center_sqr_config {
> + __u32 x_sqr_reset;
> + __u32 y_sqr_reset;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_bnr_static_config - BNR static config
> + *
> + * @wb_gains: white balance gains &ipu3_uapi_bnr_static_config_wb_gains_config
> + * @wb_gains_thr: white balance gains threshold as defined by
> + * &ipu3_uapi_bnr_static_config_wb_gains_thr_config
> + * @thr_coeffs: coefficients of threshold
> + * &ipu3_uapi_bnr_static_config_thr_coeffs_config
> + * @thr_ctrl_shd: control of shading threshold
> + * &ipu3_uapi_bnr_static_config_thr_ctrl_shd_config
> + * @opt_center: optical center &ipu3_uapi_bnr_static_config_opt_center_config
> + *
> + * Above parameters and opt_center_sqr are used for white balance and shading.
> + *
> + * @lut: lookup table &ipu3_uapi_bnr_static_config_lut_config
> + * @bp_ctrl: detect and remove bad pixels as defined in struct
> + * &ipu3_uapi_bnr_static_config_bp_ctrl_config
> + * @dn_detect_ctrl: detect and remove noise.
> + * &ipu3_uapi_bnr_static_config_dn_detect_ctrl_config
> + * @column_size: The number of pixels in column.
> + * @opt_center_sqr: Reset value of r^2 to optical center, see
> + * &ipu3_uapi_bnr_static_config_opt_center_sqr_config.
> + */
> +struct ipu3_uapi_bnr_static_config {
> + struct ipu3_uapi_bnr_static_config_wb_gains_config wb_gains;
> + struct ipu3_uapi_bnr_static_config_wb_gains_thr_config wb_gains_thr;
> + struct ipu3_uapi_bnr_static_config_thr_coeffs_config thr_coeffs;
> + struct ipu3_uapi_bnr_static_config_thr_ctrl_shd_config thr_ctrl_shd;
> + struct ipu3_uapi_bnr_static_config_opt_center_config opt_center;
> + struct ipu3_uapi_bnr_static_config_lut_config lut;
> + struct ipu3_uapi_bnr_static_config_bp_ctrl_config bp_ctrl;
> + struct ipu3_uapi_bnr_static_config_dn_detect_ctrl_config dn_detect_ctrl;
> + __u32 column_size;
> + struct ipu3_uapi_bnr_static_config_opt_center_sqr_config opt_center_sqr;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_bnr_static_config_green_disparity - Correct green disparity
> + *
> + * @gd_red: Shading gain coeff for gr disparity level in bright red region.
> + * Precision u0.6, default 4(0.0625).
> + * @reserved0: reserved
> + * @gd_green: Shading gain coeff for gr disparity level in bright green
> + * region. Precision u0.6, default 4(0.0625).
> + * @reserved1: reserved
> + * @gd_blue: Shading gain coeff for gr disparity level in bright blue region.
> + * Precision u0.6, default 4(0.0625).
> + * @reserved2: reserved
> + * @gd_black: Maximal green disparity level in dark region (stronger disparity
> + * assumed to be image detail). Precision u14, default 80.
> + * @reserved3: reserved
> + * @gd_shading: Change maximal green disparity level according to square
> + * distance from image center.
> + * @reserved4: reserved
> + * @gd_support: Lower bound for the number of second green color pixels in
> + * current pixel neighborhood with less than threshold difference
> + * from it.
> + *
> + * The shading gain coeff of red, green, blue and black are used to calculate
> + * threshold given a pixel's color value and its coordinates in the image.
> + *
> + * @reserved5: reserved
> + * @gd_clip: Turn green disparity clip on/off, [0, 1], default 1.
> + * @gd_central_weight: Central pixel weight in 9 pixels weighted sum.
> + */
> +struct ipu3_uapi_bnr_static_config_green_disparity {
> + __u32 gd_red:6;
> + __u32 reserved0:2;
> + __u32 gd_green:6;
> + __u32 reserved1:2;
> + __u32 gd_blue:6;
> + __u32 reserved2:10;
> + __u32 gd_black:14;
> + __u32 reserved3:2;
> + __u32 gd_shading:7;
> + __u32 reserved4:1;
> + __u32 gd_support:2;
> + __u32 reserved5:1;
> + __u32 gd_clip:1;
> + __u32 gd_central_weight:4;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_dm_config - De-mosaic parameters
> + *
> + * @dm_en: de-mosaic enable.
> + * @ch_ar_en: Checker artifacts removal enable flag. Default 0.
> + * @fcc_en: False color correction (FCC) enable flag. Default 0.
> + * @reserved0: reserved
> + * @frame_width: do not care
> + * @gamma_sc: Sharpening coefficient (coefficient of 2-d derivation of
> + * complementary color in Hamilton-Adams interpolation).
> + * u5, range [0, 31], default 8.
> + * @reserved1: reserved
> + * @lc_ctrl: Parameter that controls weights of Chroma Homogeneity metric
> + * in calculation of final homogeneity metric.
> + * u5, range [0, 31], default 7.
> + * @reserved2: reserved
> + * @cr_param1: First parameter that defines Checker artifact removal
> + * feature gain. Precision u5, range [0, 31], default 8.
> + * @reserved3: reserved
> + * @cr_param2: Second parameter that defines Checker artifact removal
> + * feature gain. Precision u5, range [0, 31], default 8.
> + * @reserved4: reserved
> + * @coring_param: Defines power of false color correction operation.
> + * low for preserving edge colors, high for preserving gray
> + * edge artifacts.
> + * Precision u1.4, range [0, 1.9375], default 4 (0.25).
> + * @reserved5: reserved
> + *
> + * The demosaic fixed function block is responsible to covert Bayer(mosaiced)
> + * images into color images based on demosaicing algorithm.
> + */
> +struct ipu3_uapi_dm_config {
> + __u32 dm_en:1;
> + __u32 ch_ar_en:1;
> + __u32 fcc_en:1;
> + __u32 reserved0:13;
> + __u32 frame_width:16;
> +
> + __u32 gamma_sc:5;
> + __u32 reserved1:3;
> + __u32 lc_ctrl:5;
> + __u32 reserved2:3;
> + __u32 cr_param1:5;
> + __u32 reserved3:3;
> + __u32 cr_param2:5;
> + __u32 reserved4:3;
> +
> + __u32 coring_param:5;
> + __u32 reserved5:27;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_ccm_mat_config - Color correction matrix
> + *
> + * @coeff_m11: CCM 3x3 coefficient, range [-65536, 65535]
> + * @coeff_m12: CCM 3x3 coefficient, range [-8192, 8191]
> + * @coeff_m13: CCM 3x3 coefficient, range [-32768, 32767]
> + * @coeff_o_r: Bias 3x1 coefficient, range [-8191, 8181]
> + * @coeff_m21: CCM 3x3 coefficient, range [-32767, 32767]
> + * @coeff_m22: CCM 3x3 coefficient, range [-8192, 8191]
> + * @coeff_m23: CCM 3x3 coefficient, range [-32768, 32767]
> + * @coeff_o_g: Bias 3x1 coefficient, range [-8191, 8181]
> + * @coeff_m31: CCM 3x3 coefficient, range [-32768, 32767]
> + * @coeff_m32: CCM 3x3 coefficient, range [-8192, 8191]
> + * @coeff_m33: CCM 3x3 coefficient, range [-32768, 32767]
> + * @coeff_o_b: Bias 3x1 coefficient, range [-8191, 8181]
> + *
> + * Transform sensor specific color space to standard sRGB by applying 3x3 matrix
> + * and adding a bias vector O. The transformation is basically a rotation and
> + * translation in the 3-dimensional color spaces. Here are the defaults:
> + *
> + * 9775, -2671, 1087, 0
> + * -1071, 8303, 815, 0
> + * -23, -7887, 16103, 0
> + */
> +struct ipu3_uapi_ccm_mat_config {
> + __s16 coeff_m11;
> + __s16 coeff_m12;
> + __s16 coeff_m13;
> + __s16 coeff_o_r;
> + __s16 coeff_m21;
> + __s16 coeff_m22;
> + __s16 coeff_m23;
> + __s16 coeff_o_g;
> + __s16 coeff_m31;
> + __s16 coeff_m32;
> + __s16 coeff_m33;
> + __s16 coeff_o_b;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_gamma_corr_ctrl - Gamma correction
> + *
> + * @enable: gamma correction enable.
> + * @reserved: reserved
> + */
> +struct ipu3_uapi_gamma_corr_ctrl {
> + __u32 enable:1;
> + __u32 reserved:31;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_gamma_corr_lut - Per-pixel tone mapping implemented as LUT.
> + *
> + * @lut: 256 tabulated values of the gamma function. LUT[1].. LUT[256]
> + * format u13.0, range [0, 8191].
> + *
> + * The tone mapping operation is done by a Piece wise linear graph
> + * that is implemented as a lookup table(LUT). The pixel component input
> + * intensity is the X-axis of the graph which is the table entry.
> + */
> +struct ipu3_uapi_gamma_corr_lut {
> + __u16 lut[IPU3_UAPI_GAMMA_CORR_LUT_ENTRIES];
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_gamma_config - Gamma config
> + *
> + * @gc_ctrl: control of gamma correction &ipu3_uapi_gamma_corr_ctrl
> + * @gc_lut: lookup table of gamma correction &ipu3_uapi_gamma_corr_lut
> + */
> +struct ipu3_uapi_gamma_config {
> + struct ipu3_uapi_gamma_corr_ctrl gc_ctrl __attribute__((aligned(32)));
> + struct ipu3_uapi_gamma_corr_lut gc_lut __attribute__((aligned(32)));
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_csc_mat_config - Color space conversion matrix config
> + *
> + * @coeff_c11: Conversion matrix value, format s0.14, range [-16384, 16383].
> + * @coeff_c12: Conversion matrix value, format s0.14, range [-8192, 8191].
> + * @coeff_c13: Conversion matrix value, format s0.14, range [-16384, 16383].
> + * @coeff_b1: Bias 3x1 coefficient, s13.0 range [-8192, 8191].
> + * @coeff_c21: Conversion matrix value, format s0.14, range [-16384, 16383].
> + * @coeff_c22: Conversion matrix value, format s0.14, range [-8192, 8191].
> + * @coeff_c23: Conversion matrix value, format s0.14, range [-16384, 16383].
> + * @coeff_b2: Bias 3x1 coefficient, s13.0 range [-8192, 8191].
> + * @coeff_c31: Conversion matrix value, format s0.14, range [-16384, 16383].
> + * @coeff_c32: Conversion matrix value, format s0.14, range [-8192, 8191].
> + * @coeff_c33: Conversion matrix value, format s0.14, range [-16384, 16383].
> + * @coeff_b3: Bias 3x1 coefficient, s13.0 range [-8192, 8191].
> + *
> + * To transform each pixel from RGB to YUV (Y - brightness/luminance,
> + * UV -chroma) by applying the pixel's values by a 3x3 matrix and adding an
> + * optional bias 3x1 vector. Here are the default values for the matrix:
> + *
> + * 4898, 9617, 1867, 0,
> + * -2410, -4732, 7143, 0,
> + * 10076, -8437, -1638, 0,
> + *
> + * (i.e. for real number 0.299, 0.299 * 2^14 becomes 4898.)
> + */
> +struct ipu3_uapi_csc_mat_config {
> + __s16 coeff_c11;
> + __s16 coeff_c12;
> + __s16 coeff_c13;
> + __s16 coeff_b1;
> + __s16 coeff_c21;
> + __s16 coeff_c22;
> + __s16 coeff_c23;
> + __s16 coeff_b2;
> + __s16 coeff_c31;
> + __s16 coeff_c32;
> + __s16 coeff_c33;
> + __s16 coeff_b3;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_cds_params - Chroma down-scaling
> + *
> + * @ds_c00: range [0, 3]
> + * @ds_c01: range [0, 3]
> + * @ds_c02: range [0, 3]
> + * @ds_c03: range [0, 3]
> + * @ds_c10: range [0, 3]
> + * @ds_c11: range [0, 3]
> + * @ds_c12: range [0, 3]
> + * @ds_c13: range [0, 3]
> + *
> + * In case user does not provide, above 4x2 filter will use following defaults:
> + * 1, 3, 3, 1,
> + * 1, 3, 3, 1,
> + *
> + * @ds_nf: Normalization factor for Chroma output downscaling filter,
> + * range 0,4, default 2.
> + * @reserved0: reserved
> + * @csc_en: Color space conversion enable
> + * @uv_bin_output: 0: output YUV 4.2.0, 1: output YUV 4.2.2(default).
> + * @reserved1: reserved
> + */
> +struct ipu3_uapi_cds_params {
> + __u32 ds_c00:2;
> + __u32 ds_c01:2;
> + __u32 ds_c02:2;
> + __u32 ds_c03:2;
> + __u32 ds_c10:2;
> + __u32 ds_c11:2;
> + __u32 ds_c12:2;
> + __u32 ds_c13:2;
> + __u32 ds_nf:5;
> + __u32 reserved0:3;
> + __u32 csc_en:1;
> + __u32 uv_bin_output:1;
> + __u32 reserved1:6;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_shd_grid_config - Bayer shading(darkening) correction
> + *
> + * @width: Grid horizontal dimensions, u8, [8, 128], default 73
> + * @height: Grid vertical dimensions, u8, [8, 128], default 56
> + * @block_width_log2: Log2 of the width of the grid cell in pixel count
> + * u4, [0, 15], default value 5.
> + * @reserved0: reserved
> + * @block_height_log2: Log2 of the height of the grid cell in pixel count
> + * u4, [0, 15], default value 6.
> + * @reserved1: reserved
> + * @grid_height_per_slice: SHD_MAX_CELLS_PER_SET/width.
> + * (with SHD_MAX_CELLS_PER_SET = 146).
> + * @x_start: X value of top left corner of sensor relative to ROI
> + * s13, [-4096, 0], default 0, only negative values.
> + * @y_start: Y value of top left corner of sensor relative to ROI
> + * s13, [-4096, 0], default 0, only negative values.
> + */
> +struct ipu3_uapi_shd_grid_config {
> + /* reg 0 */
> + __u8 width;
> + __u8 height;
> + __u8 block_width_log2:3;
> + __u8 reserved0:1;
> + __u8 block_height_log2:3;
> + __u8 reserved1:1;
> + __u8 grid_height_per_slice;
> + /* reg 1 */
> + __s16 x_start;
> + __s16 y_start;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_shd_general_config - Shading general config
> + *
> + * @init_set_vrt_offst_ul: set vertical offset,
> + * y_start >> block_height_log2 % grid_height_per_slice.
> + * @shd_enable: shading enable.
> + * @gain_factor: Gain factor. Shift calculated anti shading value. Precision u2.
> + * 0x0 - gain factor [1, 5], means no shift interpolated value.
> + * 0x1 - gain factor [1, 9], means shift interpolated by 1.
> + * 0x2 - gain factor [1, 17], means shift interpolated by 2.
> + * @reserved: reserved
> + *
> + * Correction is performed by multiplying a gain factor for each of the 4 Bayer
> + * channels as a function of the pixel location in the sensor.
> + */
> +struct ipu3_uapi_shd_general_config {
> + __u32 init_set_vrt_offst_ul:8;
> + __u32 shd_enable:1;
> + __u32 gain_factor:2;
> + __u32 reserved:21;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_shd_black_level_config - Black level correction
> + *
> + * @bl_r: Bios values for green red. s11 range [-2048, 2047].
> + * @bl_gr: Bios values for green blue. s11 range [-2048, 2047].
> + * @bl_gb: Bios values for red. s11 range [-2048, 2047].
> + * @bl_b: Bios values for blue. s11 range [-2048, 2047].
> + */
> +struct ipu3_uapi_shd_black_level_config {
> + __s16 bl_r;
> + __s16 bl_gr;
> + __s16 bl_gb;
> + __s16 bl_b;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_shd_config_static - Shading config static
> + *
> + * @grid: shading grid config &ipu3_uapi_shd_grid_config
> + * @general: shading general config &ipu3_uapi_shd_general_config
> + * @black_level: black level config for shading correction as defined by
> + * &ipu3_uapi_shd_black_level_config
> + */
> +struct ipu3_uapi_shd_config_static {
> + struct ipu3_uapi_shd_grid_config grid;
> + struct ipu3_uapi_shd_general_config general;
> + struct ipu3_uapi_shd_black_level_config black_level;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_shd_lut - Shading gain factor lookup table.
> + *
> + * @sets: array
> + * @sets.r_and_gr: Red and GreenR Lookup table.
> + * @sets.r_and_gr.r: Red shading factor.
> + * @sets.r_and_gr.gr: GreenR shading factor.
> + * @sets.reserved1: reserved
> + * @sets.gb_and_b: GreenB and Blue Lookup table.
> + * @sets.gb_and_b.gb: GreenB shading factor.
> + * @sets.gb_and_b.b: Blue shading factor.
> + * @sets.reserved2: reserved
> + *
> + * Map to shading correction LUT register set.
> + */
> +struct ipu3_uapi_shd_lut {
> + struct {
> + struct {
> + __u16 r;
> + __u16 gr;
> + } r_and_gr[IPU3_UAPI_SHD_MAX_CELLS_PER_SET];
> + __u8 reserved1[24];
> + struct {
> + __u16 gb;
> + __u16 b;
> + } gb_and_b[IPU3_UAPI_SHD_MAX_CELLS_PER_SET];
> + __u8 reserved2[24];
> + } sets[IPU3_UAPI_SHD_MAX_CFG_SETS];
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_shd_config - Shading config
> + *
> + * @shd: shading static config, see &ipu3_uapi_shd_config_static
> + * @shd_lut: shading lookup table &ipu3_uapi_shd_lut
> + */
> +struct ipu3_uapi_shd_config {
> + struct ipu3_uapi_shd_config_static shd __attribute__((aligned(32)));
> + struct ipu3_uapi_shd_lut shd_lut __attribute__((aligned(32)));
> +} __attribute__((packed));
> +
> +/* Image Enhancement Filter directed */
> +
> +/**
> + * struct ipu3_uapi_iefd_cux2 - IEFd Config Unit 2 parameters
> + *
> + * @x0: X0 point of Config Unit, u9.0, default 0.
> + * @x1: X1 point of Config Unit, u9.0, default 0.
> + * @a01: Slope A of Config Unit, s4.4, default 0.
> + * @b01: Slope B, always 0.
> + *
> + * Calculate weight for blending directed and non-directed denoise elements
> + *
> + * Note:
> + * Each instance of Config Unit needs X coordinate of n points and
> + * slope A factor between points calculated by driver based on calibration
> + * parameters.
> + *
> + * All CU inputs are unsigned, they will be converted to signed when written
> + * to register, i.e. a01 will be written to 9 bit register in s4.4 format.
> + * The data precision s4.4 means 4 bits for integer parts and 4 bits for the
> + * fractional part, the first bit indicates positive or negative value.
> + * For userspace software (commonly the imaging library), the computation for
> + * the CU slope values should be based on the slope resolution 1/16 (binary
> + * 0.0001 - the minimal interval value), the slope value range is [-256, +255].
> + * This applies to &ipu3_uapi_iefd_cux6_ed, &ipu3_uapi_iefd_cux2_1,
> + * &ipu3_uapi_iefd_cux2_1, &ipu3_uapi_iefd_cux4 and &ipu3_uapi_iefd_cux6_rad.
> + */
> +struct ipu3_uapi_iefd_cux2 {
> + __u32 x0:9;
> + __u32 x1:9;
> + __u32 a01:9;
> + __u32 b01:5;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_iefd_cux6_ed - Calculate power of non-directed sharpening
> + * element, Config Unit 6 for edge detail (ED).
> + *
> + * @x0: X coordinate of point 0, u9.0, default 0.
> + * @x1: X coordinate of point 1, u9.0, default 0.
> + * @x2: X coordinate of point 2, u9.0, default 0.
> + * @reserved0: reserved
> + * @x3: X coordinate of point 3, u9.0, default 0.
> + * @x4: X coordinate of point 4, u9.0, default 0.
> + * @x5: X coordinate of point 5, u9.0, default 0.
> + * @reserved1: reserved
> + * @a01: slope A points 01, s4.4, default 0.
> + * @a12: slope A points 12, s4.4, default 0.
> + * @a23: slope A points 23, s4.4, default 0.
> + * @reserved2: reserved
> + * @a34: slope A points 34, s4.4, default 0.
> + * @a45: slope A points 45, s4.4, default 0.
> + * @reserved3: reserved
> + * @b01: slope B points 01, s4.4, default 0.
> + * @b12: slope B points 12, s4.4, default 0.
> + * @b23: slope B points 23, s4.4, default 0.
> + * @reserved4: reserved
> + * @b34: slope B points 34, s4.4, default 0.
> + * @b45: slope B points 45, s4.4, default 0.
> + * @reserved5: reserved.
> + */
> +struct ipu3_uapi_iefd_cux6_ed {
> + __u32 x0:9;
> + __u32 x1:9;
> + __u32 x2:9;
> + __u32 reserved0:5;
> +
> + __u32 x3:9;
> + __u32 x4:9;
> + __u32 x5:9;
> + __u32 reserved1:5;
> +
> + __u32 a01:9;
> + __u32 a12:9;
> + __u32 a23:9;
> + __u32 reserved2:5;
> +
> + __u32 a34:9;
> + __u32 a45:9;
> + __u32 reserved3:14;
> +
> + __u32 b01:9;
> + __u32 b12:9;
> + __u32 b23:9;
> + __u32 reserved4:5;
> +
> + __u32 b34:9;
> + __u32 b45:9;
> + __u32 reserved5:14;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_iefd_cux2_1 - Calculate power of non-directed denoise
> + * element apply.
> + * @x0: X0 point of Config Unit, u9.0, default 0.
> + * @x1: X1 point of Config Unit, u9.0, default 0.
> + * @a01: Slope A of Config Unit, s4.4, default 0.
> + * @reserved1: reserved
> + * @b01: offset B0 of Config Unit, u7.0, default 0.
> + * @reserved2: reserved
> + */
> +struct ipu3_uapi_iefd_cux2_1 {
> + __u32 x0:9;
> + __u32 x1:9;
> + __u32 a01:9;
> + __u32 reserved1:5;
> +
> + __u32 b01:8;
> + __u32 reserved2:24;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_iefd_cux4 - Calculate power of non-directed sharpening
> + * element.
> + *
> + * @x0: X0 point of Config Unit, u9.0, default 0.
> + * @x1: X1 point of Config Unit, u9.0, default 0.
> + * @x2: X2 point of Config Unit, u9.0, default 0.
> + * @reserved0: reserved
> + * @x3: X3 point of Config Unit, u9.0, default 0.
> + * @a01: Slope A0 of Config Unit, s4.4, default 0.
> + * @a12: Slope A1 of Config Unit, s4.4, default 0.
> + * @reserved1: reserved
> + * @a23: Slope A2 of Config Unit, s4.4, default 0.
> + * @b01: Offset B0 of Config Unit, s7.0, default 0.
> + * @b12: Offset B1 of Config Unit, s7.0, default 0.
> + * @reserved2: reserved
> + * @b23: Offset B2 of Config Unit, s7.0, default 0.
> + * @reserved3: reserved
> + */
> +struct ipu3_uapi_iefd_cux4 {
> + __u32 x0:9;
> + __u32 x1:9;
> + __u32 x2:9;
> + __u32 reserved0:5;
> +
> + __u32 x3:9;
> + __u32 a01:9;
> + __u32 a12:9;
> + __u32 reserved1:5;
> +
> + __u32 a23:9;
> + __u32 b01:8;
> + __u32 b12:8;
> + __u32 reserved2:7;
> +
> + __u32 b23:8;
> + __u32 reserved3:24;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_iefd_cux6_rad - Radial Config Unit (CU)
> + *
> + * @x0: x0 points of Config Unit radial, u8.0
> + * @x1: x1 points of Config Unit radial, u8.0
> + * @x2: x2 points of Config Unit radial, u8.0
> + * @x3: x3 points of Config Unit radial, u8.0
> + * @x4: x4 points of Config Unit radial, u8.0
> + * @x5: x5 points of Config Unit radial, u8.0
> + * @reserved1: reserved
> + * @a01: Slope A of Config Unit radial, s7.8
> + * @a12: Slope A of Config Unit radial, s7.8
> + * @a23: Slope A of Config Unit radial, s7.8
> + * @a34: Slope A of Config Unit radial, s7.8
> + * @a45: Slope A of Config Unit radial, s7.8
> + * @reserved2: reserved
> + * @b01: Slope B of Config Unit radial, s9.0
> + * @b12: Slope B of Config Unit radial, s9.0
> + * @b23: Slope B of Config Unit radial, s9.0
> + * @reserved4: reserved
> + * @b34: Slope B of Config Unit radial, s9.0
> + * @b45: Slope B of Config Unit radial, s9.0
> + * @reserved5: reserved
> + */
> +struct ipu3_uapi_iefd_cux6_rad {
> + __u32 x0:8;
> + __u32 x1:8;
> + __u32 x2:8;
> + __u32 x3:8;
> +
> + __u32 x4:8;
> + __u32 x5:8;
> + __u32 reserved1:16;
> +
> + __u32 a01:16;
> + __u32 a12:16;
> +
> + __u32 a23:16;
> + __u32 a34:16;
> +
> + __u32 a45:16;
> + __u32 reserved2:16;
> +
> + __u32 b01:10;
> + __u32 b12:10;
> + __u32 b23:10;
> + __u32 reserved4:2;
> +
> + __u32 b34:10;
> + __u32 b45:10;
> + __u32 reserved5:12;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_iefd_cfg_units - IEFd Config Units parameters
> + *
> + * @cu_1: calculate weight for blending directed and
> + * non-directed denoise elements. See &ipu3_uapi_iefd_cux2
> + * @cu_ed: calculate power of non-directed sharpening element, see
> + * &ipu3_uapi_iefd_cux6_ed
> + * @cu_3: calculate weight for blending directed and
> + * non-directed denoise elements. A &ipu3_uapi_iefd_cux2
> + * @cu_5: calculate power of non-directed denoise element apply, use
> + * &ipu3_uapi_iefd_cux2_1
> + * @cu_6: calculate power of non-directed sharpening element. See
> + * &ipu3_uapi_iefd_cux4
> + * @cu_7: calculate weight for blending directed and
> + * non-directed denoise elements. Use &ipu3_uapi_iefd_cux2
> + * @cu_unsharp: Config Unit of unsharp &ipu3_uapi_iefd_cux4
> + * @cu_radial: Config Unit of radial &ipu3_uapi_iefd_cux6_rad
> + * @cu_vssnlm: Config Unit of vssnlm &ipu3_uapi_iefd_cux2
> + */
> +struct ipu3_uapi_yuvp1_iefd_cfg_units {
> + struct ipu3_uapi_iefd_cux2 cu_1;
> + struct ipu3_uapi_iefd_cux6_ed cu_ed;
> + struct ipu3_uapi_iefd_cux2 cu_3;
> + struct ipu3_uapi_iefd_cux2_1 cu_5;
> + struct ipu3_uapi_iefd_cux4 cu_6;
> + struct ipu3_uapi_iefd_cux2 cu_7;
> + struct ipu3_uapi_iefd_cux4 cu_unsharp;
> + struct ipu3_uapi_iefd_cux6_rad cu_radial;
> + struct ipu3_uapi_iefd_cux2 cu_vssnlm;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_iefd_config_s - IEFd config
> + *
> + * @horver_diag_coeff: Gradient compensation. Compared with vertical /
> + * horizontal (0 / 90 degree), coefficient of diagonal (45 /
> + * 135 degree) direction should be corrected by approx.
> + * 1/sqrt(2).
> + * @reserved0: reserved
> + * @clamp_stitch: Slope to stitch between clamped and unclamped edge values
> + * @reserved1: reserved
> + * @direct_metric_update: Update coeff for direction metric
> + * @reserved2: reserved
> + * @ed_horver_diag_coeff: Radial Coefficient that compensates for
> + * different distance for vertical/horizontal and
> + * diagonal gradient calculation (approx. 1/sqrt(2))
> + * @reserved3: reserved
> + */
> +struct ipu3_uapi_yuvp1_iefd_config_s {
> + __u32 horver_diag_coeff:7;
> + __u32 reserved0:1;
> + __u32 clamp_stitch:6;
> + __u32 reserved1:2;
> + __u32 direct_metric_update:5;
> + __u32 reserved2:3;
> + __u32 ed_horver_diag_coeff:7;
> + __u32 reserved3:1;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_iefd_control - IEFd control
> + *
> + * @iefd_en: Enable IEFd
> + * @denoise_en: Enable denoise
> + * @direct_smooth_en: Enable directional smooth
> + * @rad_en: Enable radial update
> + * @vssnlm_en: Enable VSSNLM output filter
> + * @reserved: reserved
> + */
> +struct ipu3_uapi_yuvp1_iefd_control {
> + __u32 iefd_en:1;
> + __u32 denoise_en:1;
> + __u32 direct_smooth_en:1;
> + __u32 rad_en:1;
> + __u32 vssnlm_en:1;
> + __u32 reserved:27;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_sharp_cfg - Sharpening config
> + *
> + * @nega_lmt_txt: Sharpening limit for negative overshoots for texture.
> + * @reserved0: reserved
> + * @posi_lmt_txt: Sharpening limit for positive overshoots for texture.
> + * @reserved1: reserved
> + * @nega_lmt_dir: Sharpening limit for negative overshoots for direction (edge).
> + * @reserved2: reserved
> + * @posi_lmt_dir: Sharpening limit for positive overshoots for direction (edge).
> + * @reserved3: reserved
> + *
> + * Fixed point type u13.0, range [0, 8191].
> + */
> +struct ipu3_uapi_sharp_cfg {
> + __u32 nega_lmt_txt:13;
> + __u32 reserved0:19;
> + __u32 posi_lmt_txt:13;
> + __u32 reserved1:19;
> + __u32 nega_lmt_dir:13;
> + __u32 reserved2:19;
> + __u32 posi_lmt_dir:13;
> + __u32 reserved3:19;
> +} __attribute__((packed));
> +
> +/**
> + * struct struct ipu3_uapi_far_w - Sharpening config for far sub-group
> + *
> + * @dir_shrp: Weight of wide direct sharpening, u1.6, range [0, 64], default 64.
> + * @reserved0: reserved
> + * @dir_dns: Weight of wide direct denoising, u1.6, range [0, 64], default 0.
> + * @reserved1: reserved
> + * @ndir_dns_powr: Power of non-direct denoising,
> + * Precision u1.6, range [0, 64], default 64.
> + * @reserved2: reserved
> + */
> +struct ipu3_uapi_far_w {
> + __u32 dir_shrp:7;
> + __u32 reserved0:1;
> + __u32 dir_dns:7;
> + __u32 reserved1:1;
> + __u32 ndir_dns_powr:7;
> + __u32 reserved2:9;
> +} __attribute__((packed));
> +
> +/**
> + * struct struct ipu3_uapi_unsharp_cfg - Unsharp config
> + *
> + * @unsharp_weight: Unsharp mask blending weight.
> + * u1.6, range [0, 64], default 16.
> + * 0 - disabled, 64 - use only unsharp.
> + * @reserved0: reserved
> + * @unsharp_amount: Unsharp mask amount, u4.5, range [0, 511], default 0.
> + * @reserved1: reserved
> + */
> +struct ipu3_uapi_unsharp_cfg {
> + __u32 unsharp_weight:7;
> + __u32 reserved0:1;
> + __u32 unsharp_amount:9;
> + __u32 reserved1:15;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_iefd_shrp_cfg - IEFd sharpness config
> + *
> + * @cfg: sharpness config &ipu3_uapi_sharp_cfg
> + * @far_w: wide range config, value as specified by &ipu3_uapi_far_w:
> + * The 5x5 environment is separated into 2 sub-groups, the 3x3 nearest
> + * neighbors (8 pixels called Near), and the second order neighborhood
> + * around them (16 pixels called Far).
> + * @unshrp_cfg: unsharpness config. &ipu3_uapi_unsharp_cfg
> + */
> +struct ipu3_uapi_yuvp1_iefd_shrp_cfg {
> + struct ipu3_uapi_sharp_cfg cfg;
> + struct ipu3_uapi_far_w far_w;
> + struct ipu3_uapi_unsharp_cfg unshrp_cfg;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_unsharp_coef0 - Unsharp mask coefficients
> + *
> + * @c00: Coeff11, s0.8, range [-255, 255], default 1.
> + * @c01: Coeff12, s0.8, range [-255, 255], default 5.
> + * @c02: Coeff13, s0.8, range [-255, 255], default 9.
> + * @reserved: reserved
> + *
> + * Configurable registers for common sharpening support.
> + */
> +struct ipu3_uapi_unsharp_coef0 {
> + __u32 c00:9;
> + __u32 c01:9;
> + __u32 c02:9;
> + __u32 reserved:5;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_unsharp_coef1 - Unsharp mask coefficients
> + *
> + * @c11: Coeff22, s0.8, range [-255, 255], default 29.
> + * @c12: Coeff23, s0.8, range [-255, 255], default 55.
> + * @c22: Coeff33, s0.8, range [-255, 255], default 96.
> + * @reserved: reserved
> + */
> +struct ipu3_uapi_unsharp_coef1 {
> + __u32 c11:9;
> + __u32 c12:9;
> + __u32 c22:9;
> + __u32 reserved:5;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_iefd_unshrp_cfg - Unsharp mask config
> + *
> + * @unsharp_coef0: unsharp coefficient 0 config. See &ipu3_uapi_unsharp_coef0
> + * @unsharp_coef1: unsharp coefficient 1 config. See &ipu3_uapi_unsharp_coef1
> + */
> +struct ipu3_uapi_yuvp1_iefd_unshrp_cfg {
> + struct ipu3_uapi_unsharp_coef0 unsharp_coef0;
> + struct ipu3_uapi_unsharp_coef1 unsharp_coef1;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_radial_reset_xy - Radial coordinate reset
> + *
> + * @x: Radial reset of x coordinate. Precision s12, [-4095, 4095], default 0.
> + * @reserved0: reserved
> + * @y: Radial center y coordinate. Precision s12, [-4095, 4095], default 0.
> + * @reserved1: reserved
> + */
> +struct ipu3_uapi_radial_reset_xy {
> + __s32 x:13;
> + __u32 reserved0:3;
> + __s32 y:13;
> + __u32 reserved1:3;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_radial_reset_x2 - Radial X^2 reset
> + *
> + * @x2: Radial reset of x^2 coordinate. Precision u24, default 0.
> + * @reserved: reserved
> + */
> +struct ipu3_uapi_radial_reset_x2 {
> + __u32 x2:24;
> + __u32 reserved:8;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_radial_reset_y2 - Radial Y^2 reset
> + *
> + * @y2: Radial reset of y^2 coordinate. Precision u24, default 0.
> + * @reserved: reserved
> + */
> +struct ipu3_uapi_radial_reset_y2 {
> + __u32 y2:24;
> + __u32 reserved:8;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_radial_cfg - Radial config
> + *
> + * @rad_nf: Radial. R^2 normalization factor is scale down by 2^ - (15 + scale)
> + * @reserved0: reserved
> + * @rad_inv_r2: Radial R^-2 normelized to (0.5..1).
> + * Precision u7, range [0, 127].
> + * @reserved1: reserved
> + */
> +struct ipu3_uapi_radial_cfg {
> + __u32 rad_nf:4;
> + __u32 reserved0:4;
> + __u32 rad_inv_r2:7;
> + __u32 reserved1:17;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_rad_far_w - Radial FAR sub-group
> + *
> + * @rad_dir_far_sharp_w: Weight of wide direct sharpening, u1.6, range [0, 64],
> + * default 64.
> + * @rad_dir_far_dns_w: Weight of wide direct denoising, u1.6, range [0, 64],
> + * default 0.
> + * @rad_ndir_far_dns_power: power of non-direct sharpening, u1.6, range [0, 64],
> + * default 0.
> + * @reserved: reserved
> + */
> +struct ipu3_uapi_rad_far_w {
> + __u32 rad_dir_far_sharp_w:8;
> + __u32 rad_dir_far_dns_w:8;
> + __u32 rad_ndir_far_dns_power:8;
> + __u32 reserved:8;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_cu_cfg0 - Radius Config Unit cfg0 register
> + *
> + * @cu6_pow: Power of CU6. Power of non-direct sharpening, u3.4.
> + * @reserved0: reserved
> + * @cu_unsharp_pow: Power of unsharp mask, u2.4.
> + * @reserved1: reserved
> + * @rad_cu6_pow: Radial/corner CU6. Directed sharpening power, u3.4.
> + * @reserved2: reserved
> + * @rad_cu_unsharp_pow: Radial power of unsharp mask, u2.4.
> + * @reserved3: reserved
> + */
> +struct ipu3_uapi_cu_cfg0 {
> + __u32 cu6_pow:7;
> + __u32 reserved0:1;
> + __u32 cu_unsharp_pow:7;
> + __u32 reserved1:1;
> + __u32 rad_cu6_pow:7;
> + __u32 reserved2:1;
> + __u32 rad_cu_unsharp_pow:6;
> + __u32 reserved3:2;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_cu_cfg1 - Radius Config Unit cfg1 register
> + *
> + * @rad_cu6_x1: X1 point of Config Unit 6, precision u9.0.
> + * @reserved0: reserved
> + * @rad_cu_unsharp_x1: X1 point for Config Unit unsharp for radial/corner point
> + * precision u9.0.
> + * @reserved1: reserved
> + */
> +struct ipu3_uapi_cu_cfg1 {
> + __u32 rad_cu6_x1:9;
> + __u32 reserved0:1;
> + __u32 rad_cu_unsharp_x1:9;
> + __u32 reserved1:13;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_iefd_rad_cfg - IEFd parameters changed radially over
> + * the picture plane.
> + *
> + * @reset_xy: reset xy value in radial calculation. &ipu3_uapi_radial_reset_xy
> + * @reset_x2: reset x square value in radial calculation. See struct
> + * &ipu3_uapi_radial_reset_x2
> + * @reset_y2: reset y square value in radial calculation. See struct
> + * &ipu3_uapi_radial_reset_y2
> + * @cfg: radial config defined in &ipu3_uapi_radial_cfg
> + * @rad_far_w: weight for wide range radial. &ipu3_uapi_rad_far_w
> + * @cu_cfg0: configuration unit 0. See &ipu3_uapi_cu_cfg0
> + * @cu_cfg1: configuration unit 1. See &ipu3_uapi_cu_cfg1
> + */
> +struct ipu3_uapi_yuvp1_iefd_rad_cfg {
> + struct ipu3_uapi_radial_reset_xy reset_xy;
> + struct ipu3_uapi_radial_reset_x2 reset_x2;
> + struct ipu3_uapi_radial_reset_y2 reset_y2;
> + struct ipu3_uapi_radial_cfg cfg;
> + struct ipu3_uapi_rad_far_w rad_far_w;
> + struct ipu3_uapi_cu_cfg0 cu_cfg0;
> + struct ipu3_uapi_cu_cfg1 cu_cfg1;
> +} __attribute__((packed));
> +
> +/* Vssnlm - Very small scale non-local mean algorithm */
> +
> +/**
> + * struct ipu3_uapi_vss_lut_x - Vssnlm LUT x0/x1/x2
> + *
> + * @vs_x0: Vssnlm LUT x0, precision u8, range [0, 255], default 16.
> + * @vs_x1: Vssnlm LUT x1, precision u8, range [0, 255], default 32.
> + * @vs_x2: Vssnlm LUT x2, precision u8, range [0, 255], default 64.
> + * @reserved2: reserved
> + */
> +struct ipu3_uapi_vss_lut_x {
> + __u32 vs_x0:8;
> + __u32 vs_x1:8;
> + __u32 vs_x2:8;
> + __u32 reserved2:8;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_vss_lut_y - Vssnlm LUT y0/y1/y2
> + *
> + * @vs_y1: Vssnlm LUT y1, precision u4, range [0, 8], default 1.
> + * @reserved0: reserved
> + * @vs_y2: Vssnlm LUT y2, precision u4, range [0, 8], default 3.
> + * @reserved1: reserved
> + * @vs_y3: Vssnlm LUT y3, precision u4, range [0, 8], default 8.
> + * @reserved2: reserved
> + */
> +struct ipu3_uapi_vss_lut_y {
> + __u32 vs_y1:4;
> + __u32 reserved0:4;
> + __u32 vs_y2:4;
> + __u32 reserved1:4;
> + __u32 vs_y3:4;
> + __u32 reserved2:12;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_iefd_vssnlm_cf - IEFd Vssnlm Lookup table
> + *
> + * @vss_lut_x: vss lookup table. See &ipu3_uapi_vss_lut_x description
> + * @vss_lut_y: vss lookup table. See &ipu3_uapi_vss_lut_y description
> + */
> +struct ipu3_uapi_yuvp1_iefd_vssnlm_cfg {
> + struct ipu3_uapi_vss_lut_x vss_lut_x;
> + struct ipu3_uapi_vss_lut_y vss_lut_y;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_iefd_config - IEFd config
> + *
> + * @units: configuration unit setting, &ipu3_uapi_yuvp1_iefd_cfg_units
> + * @config: configuration, as defined by &ipu3_uapi_yuvp1_iefd_config_s
> + * @control: control setting, as defined by &ipu3_uapi_yuvp1_iefd_control
> + * @sharp: sharpness setting, as defined by &ipu3_uapi_yuvp1_iefd_shrp_cfg
> + * @unsharp: unsharpness setting, as defined by &ipu3_uapi_yuvp1_iefd_unshrp_cfg
> + * @rad: radial setting, as defined by &ipu3_uapi_yuvp1_iefd_rad_cfg
> + * @vsslnm: vsslnm setting, as defined by &ipu3_uapi_yuvp1_iefd_vssnlm_cfg
> + */
> +struct ipu3_uapi_yuvp1_iefd_config {
> + struct ipu3_uapi_yuvp1_iefd_cfg_units units;
> + struct ipu3_uapi_yuvp1_iefd_config_s config;
> + struct ipu3_uapi_yuvp1_iefd_control control;
> + struct ipu3_uapi_yuvp1_iefd_shrp_cfg sharp;
> + struct ipu3_uapi_yuvp1_iefd_unshrp_cfg unsharp;
> + struct ipu3_uapi_yuvp1_iefd_rad_cfg rad;
> + struct ipu3_uapi_yuvp1_iefd_vssnlm_cfg vsslnm;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_yds_config - Y Down-Sampling config
> + *
> + * @c00: range [0, 3], default 0x0
> + * @c01: range [0, 3], default 0x1
> + * @c02: range [0, 3], default 0x1
> + * @c03: range [0, 3], default 0x0
> + * @c10: range [0, 3], default 0x0
> + * @c11: range [0, 3], default 0x1
> + * @c12: range [0, 3], default 0x1
> + * @c13: range [0, 3], default 0x0
> + *
> + * Above are 4x2 filter coefficients for chroma output downscaling.
> + *
> + * @norm_factor: Normalization factor, range [0, 4], default 2
> + * 0 - divide by 1
> + * 1 - divide by 2
> + * 2 - divide by 4
> + * 3 - divide by 8
> + * 4 - divide by 16
> + * @reserved0: reserved
> + * @bin_output: Down sampling on Luma channel in two optional modes
> + * 0 - Bin output 4.2.0 (default), 1 output 4.2.2.
> + * @reserved1: reserved
> + */
> +struct ipu3_uapi_yuvp1_yds_config {
> + __u32 c00:2;
> + __u32 c01:2;
> + __u32 c02:2;
> + __u32 c03:2;
> + __u32 c10:2;
> + __u32 c11:2;
> + __u32 c12:2;
> + __u32 c13:2;
> + __u32 norm_factor:5;
> + __u32 reserved0:4;
> + __u32 bin_output:1;
> + __u32 reserved1:6;
> +} __attribute__((packed));
> +
> +/* Chroma Noise Reduction */
> +
> +/**
> + * struct ipu3_uapi_yuvp1_chnr_enable_config - Chroma noise reduction enable
> + *
> + * @enable: enable/disable chroma noise reduction
> + * @yuv_mode: 0 - YUV420, 1 - YUV422
> + * @reserved0: reserved
> + * @col_size: number of columns in the frame, max width is 2560
> + * @reserved1: reserved
> + */
> +struct ipu3_uapi_yuvp1_chnr_enable_config {
> + __u32 enable:1;
> + __u32 yuv_mode:1;
> + __u32 reserved0:14;
> + __u32 col_size:12;
> + __u32 reserved1:4;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_chnr_coring_config - Coring thresholds for UV
> + *
> + * @u: U coring level, u0.13, range [0.0, 1.0], default 0.0
> + * @reserved0: reserved
> + * @v: V coring level, u0.13, range [0.0, 1.0], default 0.0
> + * @reserved1: reserved
> + */
> +struct ipu3_uapi_yuvp1_chnr_coring_config {
> + __u32 u:13;
> + __u32 reserved0:3;
> + __u32 v:13;
> + __u32 reserved1:3;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_chnr_sense_gain_config - Chroma noise reduction gains
> + *
> + * All sensitivity gain parameters have precision u13.0, range [0, 8191].
> + *
> + * @vy: Sensitivity of horizontal edge of Y, default 100
> + * @vu: Sensitivity of horizontal edge of U, default 100
> + * @vv: Sensitivity of horizontal edge of V, default 100
> + * @reserved0: reserved
> + * @hy: Sensitivity of vertical edge of Y, default 50
> + * @hu: Sensitivity of vertical edge of U, default 50
> + * @hv: Sensitivity of vertical edge of V, default 50
> + * @reserved1: reserved
> + */
> +struct ipu3_uapi_yuvp1_chnr_sense_gain_config {
> + __u32 vy:8;
> + __u32 vu:8;
> + __u32 vv:8;
> + __u32 reserved0:8;
> +
> + __u32 hy:8;
> + __u32 hu:8;
> + __u32 hv:8;
> + __u32 reserved1:8;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_chnr_iir_fir_config - Chroma IIR/FIR filter config
> + *
> + * @fir_0h: Value of center tap in horizontal FIR, range [0, 32], default 8.
> + * @reserved0: reserved
> + * @fir_1h: Value of distance 1 in horizontal FIR, range [0, 32], default 12.
> + * @reserved1: reserved
> + * @fir_2h: Value of distance 2 tap in horizontal FIR, range [0, 32], default 0.
> + * @dalpha_clip_val: weight for previous row in IIR, range [1, 256], default 0.
> + * @reserved2: reserved
> + */
> +struct ipu3_uapi_yuvp1_chnr_iir_fir_config {
> + __u32 fir_0h:6;
> + __u32 reserved0:2;
> + __u32 fir_1h:6;
> + __u32 reserved1:2;
> + __u32 fir_2h:6;
> + __u32 dalpha_clip_val:9;
> + __u32 reserved2:1;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_chnr_config - Chroma noise reduction config
> + *
> + * @enable: chroma noise reduction enable, see
> + * &ipu3_uapi_yuvp1_chnr_enable_config
> + * @coring: coring config for chroma noise reduction, see
> + * &ipu3_uapi_yuvp1_chnr_coring_config
> + * @sense_gain: sensitivity config for chroma noise reduction, see
> + * ipu3_uapi_yuvp1_chnr_sense_gain_config
> + * @iir_fir: iir and fir config for chroma noise reduction, see
> + * ipu3_uapi_yuvp1_chnr_iir_fir_config
> + */
> +struct ipu3_uapi_yuvp1_chnr_config {
> + struct ipu3_uapi_yuvp1_chnr_enable_config enable;
> + struct ipu3_uapi_yuvp1_chnr_coring_config coring;
> + struct ipu3_uapi_yuvp1_chnr_sense_gain_config sense_gain;
> + struct ipu3_uapi_yuvp1_chnr_iir_fir_config iir_fir;
> +} __attribute__((packed));
> +
> +/* Edge Enhancement and Noise Reduction */
> +
> +/**
> + * struct ipu3_uapi_yuvp1_y_ee_nr_lpf_config - Luma(Y) edge enhancement low-pass
> + * filter coefficients
> + *
> + * @a_diag: Smoothing diagonal coefficient, u5.0.
> + * @reserved0: reserved
> + * @a_periph: Image smoothing perpherial, u5.0.
> + * @reserved1: reserved
> + * @a_cent: Image Smoothing center coefficient, u5.0.
> + * @reserved2: reserved
> + * @enable: 0: Y_EE_NR disabled, output = input; 1: Y_EE_NR enabled.
> + */
> +struct ipu3_uapi_yuvp1_y_ee_nr_lpf_config {
> + __u32 a_diag:5;
> + __u32 reserved0:3;
> + __u32 a_periph:5;
> + __u32 reserved1:3;
> + __u32 a_cent:5;
> + __u32 reserved2:9;
> + __u32 enable:1;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_y_ee_nr_sense_config - Luma(Y) edge enhancement
> + * noise reduction sensitivity gains
> + *
> + * @edge_sense_0: Sensitivity of edge in dark area. u13.0, default 8191.
> + * @reserved0: reserved
> + * @delta_edge_sense: Difference in the sensitivity of edges between
> + * the bright and dark areas. u13.0, default 0.
> + * @reserved1: reserved
> + * @corner_sense_0: Sensitivity of corner in dark area. u13.0, default 0.
> + * @reserved2: reserved
> + * @delta_corner_sense: Difference in the sensitivity of corners between
> + * the bright and dark areas. u13.0, default 8191.
> + * @reserved3: reserved
> + */
> +struct ipu3_uapi_yuvp1_y_ee_nr_sense_config {
> + __u32 edge_sense_0:13;
> + __u32 reserved0:3;
> + __u32 delta_edge_sense:13;
> + __u32 reserved1:3;
> + __u32 corner_sense_0:13;
> + __u32 reserved2:3;
> + __u32 delta_corner_sense:13;
> + __u32 reserved3:3;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_y_ee_nr_gain_config - Luma(Y) edge enhancement
> + * noise reduction gain config
> + *
> + * @gain_pos_0: Gain for positive edge in dark area. u5.0, [0, 16], default 2.
> + * @reserved0: reserved
> + * @delta_gain_posi: Difference in the gain of edges between the bright and
> + * dark areas for positive edges. u5.0, [0, 16], default 0.
> + * @reserved1: reserved
> + * @gain_neg_0: Gain for negative edge in dark area. u5.0, [0, 16], default 8.
> + * @reserved2: reserved
> + * @delta_gain_neg: Difference in the gain of edges between the bright and
> + * dark areas for negative edges. u5.0, [0, 16], default 0.
> + * @reserved3: reserved
> + */
> +struct ipu3_uapi_yuvp1_y_ee_nr_gain_config {
> + __u32 gain_pos_0:5;
> + __u32 reserved0:3;
> + __u32 delta_gain_posi:5;
> + __u32 reserved1:3;
> + __u32 gain_neg_0:5;
> + __u32 reserved2:3;
> + __u32 delta_gain_neg:5;
> + __u32 reserved3:3;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_y_ee_nr_clip_config - Luma(Y) edge enhancement
> + * noise reduction clipping config
> + *
> + * @clip_pos_0: Limit of positive edge in dark area
> + * u5, value [0, 16], default 8.
> + * @reserved0: reserved
> + * @delta_clip_posi: Difference in the limit of edges between the bright
> + * and dark areas for positive edges.
> + * u5, value [0, 16], default 8.
> + * @reserved1: reserved
> + * @clip_neg_0: Limit of negative edge in dark area
> + * u5, value [0, 16], default 8.
> + * @reserved2: reserved
> + * @delta_clip_neg: Difference in the limit of edges between the bright
> + * and dark areas for negative edges.
> + * u5, value [0, 16], default 8.
> + * @reserved3: reserved
> + */
> +struct ipu3_uapi_yuvp1_y_ee_nr_clip_config {
> + __u32 clip_pos_0:5;
> + __u32 reserved0:3;
> + __u32 delta_clip_posi:5;
> + __u32 reserved1:3;
> + __u32 clip_neg_0:5;
> + __u32 reserved2:3;
> + __u32 delta_clip_neg:5;
> + __u32 reserved3:3;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_y_ee_nr_frng_config - Luma(Y) edge enhancement
> + * noise reduction fringe config
> + *
> + * @gain_exp: Common exponent of gains, u4, [0, 8], default 2.
> + * @reserved0: reserved
> + * @min_edge: Threshold for edge and smooth stitching, u13.
> + * @reserved1: reserved
> + * @lin_seg_param: Power of LinSeg, u4.
> + * @reserved2: reserved
> + * @t1: Parameter for enabling/disabling the edge enhancement, u1.0, [0, 1],
> + * default 1.
> + * @t2: Parameter for enabling/disabling the smoothing, u1.0, [0, 1],
> + * default 1.
> + * @reserved3: reserved
> + */
> +struct ipu3_uapi_yuvp1_y_ee_nr_frng_config {
> + __u32 gain_exp:4;
> + __u32 reserved0:28;
> + __u32 min_edge:13;
> + __u32 reserved1:3;
> + __u32 lin_seg_param:4;
> + __u32 reserved2:4;
> + __u32 t1:1;
> + __u32 t2:1;
> + __u32 reserved3:6;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_y_ee_nr_diag_config - Luma(Y) edge enhancement
> + * noise reduction diagonal config
> + *
> + * @diag_disc_g: Coefficient that prioritize diagonal edge direction on
> + * horizontal or vertical for final enhancement.
> + * u4.0, [1, 15], default 1.
> + * @reserved0: reserved
> + * @hvw_hor: Weight of horizontal/vertical edge enhancement for hv edge.
> + * u2.2, [1, 15], default 4.
> + * @dw_hor: Weight of diagonal edge enhancement for hv edge.
> + * u2.2, [1, 15], default 1.
> + * @hvw_diag: Weight of horizontal/vertical edge enhancement for diagonal edge.
> + * u2.2, [1, 15], default 1.
> + * @dw_diag: Weight of diagonal edge enhancement for diagonal edge.
> + * u2.2, [1, 15], default 4.
> + * @reserved1: reserved
> + */
> +struct ipu3_uapi_yuvp1_y_ee_nr_diag_config {
> + __u32 diag_disc_g:4;
> + __u32 reserved0:4;
> + __u32 hvw_hor:4;
> + __u32 dw_hor:4;
> + __u32 hvw_diag:4;
> + __u32 dw_diag:4;
> + __u32 reserved1:8;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_y_ee_nr_fc_coring_config - Luma(Y) edge enhancement
> + * noise reduction false color correction (FCC) coring config
> + *
> + * @pos_0: Gain for positive edge in dark, u13.0, [0, 16], default 0.
> + * @reserved0: reserved
> + * @pos_delta: Gain for positive edge in bright, value: pos_0 + pos_delta <=16
> + * u13.0, default 0.
> + * @reserved1: reserved
> + * @neg_0: Gain for negative edge in dark area, u13.0, range [0, 16], default 0.
> + * @reserved2: reserved
> + * @neg_delta: Gain for negative edge in bright area. neg_0 + neg_delta <=16
> + * u13.0, default 0.
> + * @reserved3: reserved
> + *
> + * Coring is a simple soft thresholding technique.
> + */
> +struct ipu3_uapi_yuvp1_y_ee_nr_fc_coring_config {
> + __u32 pos_0:13;
> + __u32 reserved0:3;
> + __u32 pos_delta:13;
> + __u32 reserved1:3;
> + __u32 neg_0:13;
> + __u32 reserved2:3;
> + __u32 neg_delta:13;
> + __u32 reserved3:3;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp1_y_ee_nr_config - Edge enhancement and noise reduction
> + *
> + * @lpf: low-pass filter config. See &ipu3_uapi_yuvp1_y_ee_nr_lpf_config
> + * @sense: sensitivity config. See &ipu3_uapi_yuvp1_y_ee_nr_sense_config
> + * @gain: gain config as defined in &ipu3_uapi_yuvp1_y_ee_nr_gain_config
> + * @clip: clip config as defined in &ipu3_uapi_yuvp1_y_ee_nr_clip_config
> + * @frng: fringe config as defined in &ipu3_uapi_yuvp1_y_ee_nr_frng_config
> + * @diag: diagonal edge config. See &ipu3_uapi_yuvp1_y_ee_nr_diag_config
> + * @fc_coring: coring config for fringe control. See
> + * &ipu3_uapi_yuvp1_y_ee_nr_fc_coring_config
> + */
> +struct ipu3_uapi_yuvp1_y_ee_nr_config {
> + struct ipu3_uapi_yuvp1_y_ee_nr_lpf_config lpf;
> + struct ipu3_uapi_yuvp1_y_ee_nr_sense_config sense;
> + struct ipu3_uapi_yuvp1_y_ee_nr_gain_config gain;
> + struct ipu3_uapi_yuvp1_y_ee_nr_clip_config clip;
> + struct ipu3_uapi_yuvp1_y_ee_nr_frng_config frng;
> + struct ipu3_uapi_yuvp1_y_ee_nr_diag_config diag;
> + struct ipu3_uapi_yuvp1_y_ee_nr_fc_coring_config fc_coring;
> +} __attribute__((packed));
> +
> +/* Total Color Correction */
> +
> +/**
> + * struct ipu3_uapi_yuvp2_tcc_gen_control_static_config - Total color correction
> + * general control config
> + *
> + * @en: 0 - TCC disabled. Output = input 1 - TCC enabled.
> + * @blend_shift: blend shift, Range[3, 4], default NA.
> + * @gain_according_to_y_only: 0: Gain is calculated according to YUV,
> + * 1: Gain is calculated according to Y only
> + * @reserved0: reserved
> + * @gamma: Final blending coefficients. Values[-16, 16], default NA.
> + * @reserved1: reserved
> + * @delta: Final blending coefficients. Values[-16, 16], default NA.
> + * @reserved2: reserved
> + */
> +struct ipu3_uapi_yuvp2_tcc_gen_control_static_config {
> + __u32 en:1;
> + __u32 blend_shift:3;
> + __u32 gain_according_to_y_only:1;
> + __u32 reserved0:11;
> + __s32 gamma:5;
> + __u32 reserved1:3;
> + __s32 delta:5;
> + __u32 reserved2:3;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp2_tcc_macc_elem_static_config - Total color correction
> + * multi-axis color control (MACC) config
> + *
> + * @a: a coefficient for 2x2 MACC conversion matrix.
> + * @reserved0: reserved
> + * @b: b coefficient 2x2 MACC conversion matrix.
> + * @reserved1: reserved
> + * @c: c coefficient for 2x2 MACC conversion matrix.
> + * @reserved2: reserved
> + * @d: d coefficient for 2x2 MACC conversion matrix.
> + * @reserved3: reserved
> + */
> +struct ipu3_uapi_yuvp2_tcc_macc_elem_static_config {
> + __s32 a:12;
> + __u32 reserved0:4;
> + __s32 b:12;
> + __u32 reserved1:4;
> + __s32 c:12;
> + __u32 reserved2:4;
> + __s32 d:12;
> + __u32 reserved3:4;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp2_tcc_macc_table_static_config - Total color correction
> + * multi-axis color control (MACC) table array
> + *
> + * @entries: config for multi axis color correction, as specified by
> + * &ipu3_uapi_yuvp2_tcc_macc_elem_static_config
> + */
> +struct ipu3_uapi_yuvp2_tcc_macc_table_static_config {
> + struct ipu3_uapi_yuvp2_tcc_macc_elem_static_config
> + entries[IPU3_UAPI_YUVP2_TCC_MACC_TABLE_ELEMENTS];
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp2_tcc_inv_y_lut_static_config - Total color correction
> + * inverse y lookup table
> + *
> + * @entries: lookup table for inverse y estimation, and use it to estimate the
> + * ratio between luma and chroma. Chroma by approximate the absolute
> + * value of the radius on the chroma plane (R = sqrt(u^2+v^2) ) and
> + * luma by approximate by 1/Y.
> + */
> +struct ipu3_uapi_yuvp2_tcc_inv_y_lut_static_config {
> + __u16 entries[IPU3_UAPI_YUVP2_TCC_INV_Y_LUT_ELEMENTS];
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp2_tcc_gain_pcwl_lut_static_config - Total color
> + * correction lookup table for PCWL
> + *
> + * @entries: lookup table for gain piece wise linear transformation (PCWL)
> + */
> +struct ipu3_uapi_yuvp2_tcc_gain_pcwl_lut_static_config {
> + __u16 entries[IPU3_UAPI_YUVP2_TCC_GAIN_PCWL_LUT_ELEMENTS];
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp2_tcc_r_sqr_lut_static_config - Total color correction
> + * lookup table for r square root
> + *
> + * @entries: lookup table for r square root estimation
> + */
> +struct ipu3_uapi_yuvp2_tcc_r_sqr_lut_static_config {
> + __s16 entries[IPU3_UAPI_YUVP2_TCC_R_SQR_LUT_ELEMENTS];
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_yuvp2_tcc_static_config- Total color correction static
> + *
> + * @gen_control: general config for Total Color Correction
> + * @macc_table: config for multi axis color correction
> + * @inv_y_lut: lookup table for inverse y estimation
> + * @gain_pcwl: lookup table for gain PCWL
> + * @r_sqr_lut: lookup table for r square root estimation.
> + */
> +struct ipu3_uapi_yuvp2_tcc_static_config {
> + struct ipu3_uapi_yuvp2_tcc_gen_control_static_config gen_control;
> + struct ipu3_uapi_yuvp2_tcc_macc_table_static_config macc_table;
> + struct ipu3_uapi_yuvp2_tcc_inv_y_lut_static_config inv_y_lut;
> + struct ipu3_uapi_yuvp2_tcc_gain_pcwl_lut_static_config gain_pcwl;
> + struct ipu3_uapi_yuvp2_tcc_r_sqr_lut_static_config r_sqr_lut;
> +} __attribute__((packed));
> +
> +/* Advanced Noise Reduction related structs */
> +
> +/*
> + * struct ipu3_uapi_anr_alpha - Advanced noise reduction alpha
> + *
> + * Tunable parameters that are subject to modification according to the
> + * total gain used.
> + */
> +struct ipu3_uapi_anr_alpha {
> + __u16 gr;
> + __u16 r;
> + __u16 b;
> + __u16 gb;
> + __u16 dc_gr;
> + __u16 dc_r;
> + __u16 dc_b;
> + __u16 dc_gb;
> +} __attribute__((packed));
> +
> +/*
> + * struct ipu3_uapi_anr_beta - Advanced noise reduction beta
> + *
> + * Tunable parameters that are subject to modification according to the
> + * total gain used.
> + */
> +struct ipu3_uapi_anr_beta {
> + __u16 beta_gr;
> + __u16 beta_r;
> + __u16 beta_b;
> + __u16 beta_gb;
> +} __attribute__((packed));
> +
> +/*
> + * struct ipu3_uapi_anr_plane_color - Advanced noise reduction per plane R, Gr,
> + * Gb and B register settings
> + *
> + * Tunable parameters that are subject to modification according to the
> + * total gain used.
> + */
> +struct ipu3_uapi_anr_plane_color {
> + __u16 reg_w_gr[16];
> + __u16 reg_w_r[16];
> + __u16 reg_w_b[16];
> + __u16 reg_w_gb[16];
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_anr_transform_config - Advanced noise reduction transform
> + *
> + * @enable: advanced noise reduction enabled.
> + * @adaptive_treshhold_en: On IPU3, adaptive threshold is always enabled.
> + * @reserved1: reserved
> + * @reserved2: reserved
> + * @alpha: using following defaults:
> + * 13, 13, 13, 13, 0, 0, 0, 0
> + * 11, 11, 11, 11, 0, 0, 0, 0
> + * 14, 14, 14, 14, 0, 0, 0, 0
> + * @beta: use following defaults:
> + * 24, 24, 24, 24
> + * 21, 20, 20, 21
> + * 25, 25, 25, 25
> + * @color: use defaults defined in driver/media/pci/intel/ipu3-tables.c
> + * @sqrt_lut: 11 bits per element, values =
> + * [724 768 810 849 887
> + * 923 958 991 1024 1056
> + * 1116 1145 1173 1201 1086
> + * 1228 1254 1280 1305 1330
> + * 1355 1379 1402 1425 1448]
> + * @xreset: Reset value of X for r^2 calculation Value: col_start-X_center
> + * Constraint: Xreset + FrameWdith=4095 Xreset= -4095, default -1632.
> + * @reserved3: reserved
> + * @yreset: Reset value of Y for r^2 calculation Value: row_start-Y_center
> + * Constraint: Yreset + FrameHeight=4095 Yreset= -4095, default -1224.
> + * @reserved4: reserved
> + * @x_sqr_reset: Reset value of X^2 for r^2 calculation Value = (Xreset)^2
> + * @r_normfactor: Normalization factor for R. Default 14.
> + * @reserved5: reserved
> + * @y_sqr_reset: Reset value of Y^2 for r^2 calculation Value = (Yreset)^2
> + * @gain_scale: Parameter describing shading gain as a function of distance
> + * from the image center.
> + * A single value per frame, loaded by the driver. Default 115.
> + */
> +struct ipu3_uapi_anr_transform_config {
> + __u32 enable:1; /* 0 or 1, disabled or enabled */
> + __u32 adaptive_treshhold_en:1; /* On IPU3, always enabled */
> +
> + __u32 reserved1:30;
> + __u8 reserved2[44];
> +
> + struct ipu3_uapi_anr_alpha alpha[3];
> + struct ipu3_uapi_anr_beta beta[3];
> + struct ipu3_uapi_anr_plane_color color[3];
> +
> + __u16 sqrt_lut[IPU3_UAPI_ANR_LUT_SIZE]; /* 11 bits per element */
> +
> + __s16 xreset:13;
> + __u16 reserved3:3;
> + __s16 yreset:13;
> + __u16 reserved4:3;
> +
> + __u32 x_sqr_reset:24;
> + __u32 r_normfactor:5;
> + __u32 reserved5:3;
> +
> + __u32 y_sqr_reset:24;
> + __u32 gain_scale:8;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_anr_stitch_pyramid - ANR stitch pyramid
> + *
> + * @entry0: pyramid LUT entry0, range [0x0, 0x3f]
> + * @entry1: pyramid LUT entry1, range [0x0, 0x3f]
> + * @entry2: pyramid LUT entry2, range [0x0, 0x3f]
> + * @reserved: reserved
> + */
> +struct ipu3_uapi_anr_stitch_pyramid {
> + __u32 entry0:6;
> + __u32 entry1:6;
> + __u32 entry2:6;
> + __u32 reserved:14;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_anr_stitch_config - ANR stitch config
> + *
> + * @anr_stitch_en: enable stitch. Enabled with 1.
> + * @reserved: reserved
> + * @pyramid: pyramid table as defined by &ipu3_uapi_anr_stitch_pyramid
> + * default values:
> + * { 1, 3, 5 }, { 7, 7, 5 }, { 3, 1, 3 },
> + * { 9, 15, 21 }, { 21, 15, 9 }, { 3, 5, 15 },
> + * { 25, 35, 35 }, { 25, 15, 5 }, { 7, 21, 35 },
> + * { 49, 49, 35 }, { 21, 7, 7 }, { 21, 35, 49 },
> + * { 49, 35, 21 }, { 7, 5, 15 }, { 25, 35, 35 },
> + * { 25, 15, 5 }, { 3, 9, 15 }, { 21, 21, 15 },
> + * { 9, 3, 1 }, { 3, 5, 7 }, { 7, 5, 3}, { 1 }
> + */
> +struct ipu3_uapi_anr_stitch_config {
> + __u32 anr_stitch_en;
> + __u8 reserved[44];
> + struct ipu3_uapi_anr_stitch_pyramid pyramid[IPU3_UAPI_ANR_PYRAMID_SIZE];
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_anr_config - ANR config
> + *
> + * @transform: advanced noise reduction transform config as specified by
> + * &ipu3_uapi_anr_transform_config
> + * @stitch: create 4x4 patch from 4 surrounding 8x8 patches.
> + */
> +struct ipu3_uapi_anr_config {
> + struct ipu3_uapi_anr_transform_config transform __attribute__((aligned(32)));
> + struct ipu3_uapi_anr_stitch_config stitch __attribute__((aligned(32)));
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_acc_param - Accelerator cluster parameters
> + *
> + * ACC refers to the HW cluster containing all Fixed Functions (FFs). Each FF
> + * implements a specific algorithm.
> + *
> + * @bnr: parameters for bayer noise reduction static config. See
> + * &ipu3_uapi_bnr_static_config
> + * @green_disparity: disparity static config between gr and gb channel.
> + * See &ipu3_uapi_bnr_static_config_green_disparity
> + * @dm: de-mosaic config. See &ipu3_uapi_dm_config
> + * @ccm: color correction matrix. See &ipu3_uapi_ccm_mat_config
> + * @gamma: gamma correction config. See &ipu3_uapi_gamma_config
> + * @csc: color space conversion matrix. See &ipu3_uapi_csc_mat_config
> + * @cds: color down sample config. See &ipu3_uapi_cds_params
> + * @shd: lens shading correction config. See &ipu3_uapi_shd_config
> + * @iefd: Image enhancement filter and denoise config.
> + * &ipu3_uapi_yuvp1_iefd_config
> + * @yds_c0: y down scaler config. &ipu3_uapi_yuvp1_yds_config
> + * @chnr_c0: chroma noise reduction config. &ipu3_uapi_yuvp1_chnr_config
> + * @y_ee_nr: y edge enhancement and noise reduction config.
> + * &ipu3_uapi_yuvp1_y_ee_nr_config
> + * @yds: y down scaler config. See &ipu3_uapi_yuvp1_yds_config
> + * @chnr: chroma noise reduction config. See &ipu3_uapi_yuvp1_chnr_config
> + * @reserved1: reserved
> + * @yds2: y channel down scaler config. See &ipu3_uapi_yuvp1_yds_config
> + * @tcc: total color correction config as defined in struct
> + * &ipu3_uapi_yuvp2_tcc_static_config
> + * @reserved2: reserved
> + * @anr: advanced noise reduction config.See &ipu3_uapi_anr_config
> + * @awb_fr: AWB filter response config. See ipu3_uapi_awb_fr_config
> + * @ae: auto exposure config As specified by &ipu3_uapi_ae_config
> + * @af: auto focus config. As specified by &ipu3_uapi_af_config
> + * @awb: auto white balance config. As specified by &ipu3_uapi_awb_config
> + */
> +struct ipu3_uapi_acc_param {
> + struct ipu3_uapi_bnr_static_config bnr;
> + struct ipu3_uapi_bnr_static_config_green_disparity
> + green_disparity __attribute__((aligned(32)));
> + struct ipu3_uapi_dm_config dm __attribute__((aligned(32)));
> + struct ipu3_uapi_ccm_mat_config ccm __attribute__((aligned(32)));
> + struct ipu3_uapi_gamma_config gamma __attribute__((aligned(32)));
> + struct ipu3_uapi_csc_mat_config csc __attribute__((aligned(32)));
> + struct ipu3_uapi_cds_params cds __attribute__((aligned(32)));
> + struct ipu3_uapi_shd_config shd __attribute__((aligned(32)));
> + struct ipu3_uapi_yuvp1_iefd_config iefd __attribute__((aligned(32)));
> + struct ipu3_uapi_yuvp1_yds_config yds_c0 __attribute__((aligned(32)));
> + struct ipu3_uapi_yuvp1_chnr_config chnr_c0 __attribute__((aligned(32)));
> + struct ipu3_uapi_yuvp1_y_ee_nr_config y_ee_nr __attribute__((aligned(32)));
> + struct ipu3_uapi_yuvp1_yds_config yds __attribute__((aligned(32)));
> + struct ipu3_uapi_yuvp1_chnr_config chnr __attribute__((aligned(32)));
> + struct ipu3_uapi_yuvp1_yds_config yds2 __attribute__((aligned(32)));
> + struct ipu3_uapi_yuvp2_tcc_static_config tcc __attribute__((aligned(32)));
> + struct ipu3_uapi_anr_config anr;
> + struct ipu3_uapi_awb_fr_config_s awb_fr;
> + struct ipu3_uapi_ae_config ae;
> + struct ipu3_uapi_af_config_s af;
> + struct ipu3_uapi_awb_config awb;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_isp_lin_vmem_params - Linearization parameters
> + *
> + * @lin_lutlow_gr: linearization look-up table for GR channel interpolation.
> + * @lin_lutlow_r: linearization look-up table for R channel interpolation.
> + * @lin_lutlow_b: linearization look-up table for B channel interpolation.
> + * @lin_lutlow_gb: linearization look-up table for GB channel interpolation.
> + * lin_lutlow_gr / lin_lutlow_r / lin_lutlow_b /
> + * lin_lutlow_gb <= LIN_MAX_VALUE - 1.
> + * @lin_lutdif_gr: lin_lutlow_gr[i+1] - lin_lutlow_gr[i].
> + * @lin_lutdif_r: lin_lutlow_r[i+1] - lin_lutlow_r[i].
> + * @lin_lutdif_b: lin_lutlow_b[i+1] - lin_lutlow_b[i].
> + * @lin_lutdif_gb: lin_lutlow_gb[i+1] - lin_lutlow_gb[i].
> + */
> +struct ipu3_uapi_isp_lin_vmem_params {
> + __s16 lin_lutlow_gr[IPU3_UAPI_LIN_LUT_SIZE];
> + __s16 lin_lutlow_r[IPU3_UAPI_LIN_LUT_SIZE];
> + __s16 lin_lutlow_b[IPU3_UAPI_LIN_LUT_SIZE];
> + __s16 lin_lutlow_gb[IPU3_UAPI_LIN_LUT_SIZE];
> + __s16 lin_lutdif_gr[IPU3_UAPI_LIN_LUT_SIZE];
> + __s16 lin_lutdif_r[IPU3_UAPI_LIN_LUT_SIZE];
> + __s16 lin_lutdif_b[IPU3_UAPI_LIN_LUT_SIZE];
> + __s16 lin_lutdif_gb[IPU3_UAPI_LIN_LUT_SIZE];
> +} __attribute__((packed));
> +
> +/* Temporal Noise Reduction */
> +
> +/**
> + * struct ipu3_uapi_isp_tnr3_vmem_params - Temporal noise reduction vector
> + * memory parameters
> + *
> + * @slope: slope setting in interpolation curve for temporal noise reduction.
> + * @reserved1: reserved
> + * @sigma: knee point setting in interpolation curve for temporal
> + * noise reduction.
> + * @reserved2: reserved
> + */
> +struct ipu3_uapi_isp_tnr3_vmem_params {
> + __u16 slope[IPU3_UAPI_ISP_TNR3_VMEM_LEN];
> + __u16 reserved1[IPU3_UAPI_ISP_VEC_ELEMS
> + - IPU3_UAPI_ISP_TNR3_VMEM_LEN];
> + __u16 sigma[IPU3_UAPI_ISP_TNR3_VMEM_LEN];
> + __u16 reserved2[IPU3_UAPI_ISP_VEC_ELEMS
> + - IPU3_UAPI_ISP_TNR3_VMEM_LEN];
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_isp_tnr3_params - Temporal noise reduction v3 parameters
> + *
> + * @knee_y1: Knee point TNR3 assumes standard deviation of Y,U and
> + * V at Y1 are TnrY1_Sigma_Y, U and V.
> + * @knee_y2: Knee point TNR3 assumes standard deviation of Y,U and
> + * V at Y2 are TnrY2_Sigma_Y, U and V.
> + * @maxfb_y: Max feedback gain for Y
> + * @maxfb_u: Max feedback gain for U
> + * @maxfb_v: Max feedback gain for V
> + * @round_adj_y: rounding Adjust for Y
> + * @round_adj_u: rounding Adjust for U
> + * @round_adj_v: rounding Adjust for V
> + * @ref_buf_select: selection of the reference frame buffer to be used.
> + */
> +struct ipu3_uapi_isp_tnr3_params {
> + __u32 knee_y1;
> + __u32 knee_y2;
> + __u32 maxfb_y;
> + __u32 maxfb_u;
> + __u32 maxfb_v;
> + __u32 round_adj_y;
> + __u32 round_adj_u;
> + __u32 round_adj_v;
> + __u32 ref_buf_select;
> +} __attribute__((packed));
> +
> +/* Extreme Noise Reduction version 3 */
> +
> +/**
> + * struct ipu3_uapi_isp_xnr3_vmem_params - Extreme noise reduction v3
> + * vector memory parameters
> + *
> + * @x: xnr3 parameters.
> + * @a: xnr3 parameters.
> + * @b: xnr3 parameters.
> + * @c: xnr3 parameters.
> + */
> +struct ipu3_uapi_isp_xnr3_vmem_params {
> + __u16 x[IPU3_UAPI_ISP_VEC_ELEMS];
> + __u16 a[IPU3_UAPI_ISP_VEC_ELEMS];
> + __u16 b[IPU3_UAPI_ISP_VEC_ELEMS];
> + __u16 c[IPU3_UAPI_ISP_VEC_ELEMS];
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_xnr3_alpha_params - Extreme noise reduction v3
> + * alpha tuning parameters
> + *
> + * @y0: Sigma for Y range similarity in dark area.
> + * @u0: Sigma for U range similarity in dark area.
> + * @v0: Sigma for V range similarity in dark area.
> + * @ydiff: Sigma difference for Y between bright area and dark area.
> + * @udiff: Sigma difference for U between bright area and dark area.
> + * @vdiff: Sigma difference for V between bright area and dark area.
> + */
> +struct ipu3_uapi_xnr3_alpha_params {
> + __u32 y0;
> + __u32 u0;
> + __u32 v0;
> + __u32 ydiff;
> + __u32 udiff;
> + __u32 vdiff;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_xnr3_coring_params - Extreme noise reduction v3
> + * coring parameters
> + *
> + * @u0: Coring Threshold of U channel in dark area.
> + * @v0: Coring Threshold of V channel in dark area.
> + * @udiff: Threshold difference of U channel between bright and dark area.
> + * @vdiff: Threshold difference of V channel between bright and dark area.
> + */
> +struct ipu3_uapi_xnr3_coring_params {
> + __u32 u0;
> + __u32 v0;
> + __u32 udiff;
> + __u32 vdiff;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_xnr3_blending_params - Blending factor
> + *
> + * @strength: The factor for blending output with input. This is tuning
> + * parameterHigher values lead to more aggressive XNR operation.
> + */
> +struct ipu3_uapi_xnr3_blending_params {
> + __u32 strength;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_isp_xnr3_params - Extreme noise reduction v3 parameters
> + *
> + * @alpha: parameters for xnr3 alpha. See &ipu3_uapi_xnr3_alpha_params
> + * @coring: parameters for xnr3 coring. See &ipu3_uapi_xnr3_coring_params
> + * @blending: parameters for xnr3 blending. See &ipu3_uapi_xnr3_blending_params
> + */
> +struct ipu3_uapi_isp_xnr3_params {
> + struct ipu3_uapi_xnr3_alpha_params alpha;
> + struct ipu3_uapi_xnr3_coring_params coring;
> + struct ipu3_uapi_xnr3_blending_params blending;
> +} __attribute__((packed));
> +
> +/***** Obgrid (optical black level compensation) table entry *****/
> +
> +/**
> + * struct ipu3_uapi_obgrid_param - Optical black level compensation parameters
> + *
> + * @gr: Grid table values for color GR
> + * @r: Grid table values for color R
> + * @b: Grid table values for color B
> + * @gb: Grid table values for color GB
> + *
> + * Black level is different for red, green, and blue channels. So black level
> + * compensation is different per channel.
> + */
> +struct ipu3_uapi_obgrid_param {
> + __u16 gr;
> + __u16 r;
> + __u16 b;
> + __u16 gb;
> +} __attribute__((packed));
> +
> +/******************* V4L2_META_FMT_IPU3_PARAMS *******************/
> +
> +/**
> + * struct ipu3_uapi_flags - bits to indicate which pipeline needs update
> + *
> + * @gdc: 0 = no update, 1 = update.
> + * @obgrid: 0 = no update, 1 = update.
> + * @reserved1: Not used.
> + * @acc_bnr: 0 = no update, 1 = update.
> + * @acc_green_disparity: 0 = no update, 1 = update.
> + * @acc_dm: 0 = no update, 1 = update.
> + * @acc_ccm: 0 = no update, 1 = update.
> + * @acc_gamma: 0 = no update, 1 = update.
> + * @acc_csc: 0 = no update, 1 = update.
> + * @acc_cds: 0 = no update, 1 = update.
> + * @acc_shd: 0 = no update, 1 = update.
> + * @reserved2: Not used.
> + * @acc_iefd: 0 = no update, 1 = update.
> + * @acc_yds_c0: 0 = no update, 1 = update.
> + * @acc_chnr_c0: 0 = no update, 1 = update.
> + * @acc_y_ee_nr: 0 = no update, 1 = update.
> + * @acc_yds: 0 = no update, 1 = update.
> + * @acc_chnr: 0 = no update, 1 = update.
> + * @acc_ytm: 0 = no update, 1 = update.
> + * @acc_yds2: 0 = no update, 1 = update.
> + * @acc_tcc: 0 = no update, 1 = update.
> + * @acc_dpc: 0 = no update, 1 = update.
> + * @acc_bds: 0 = no update, 1 = update.
> + * @acc_anr: 0 = no update, 1 = update.
> + * @acc_awb_fr: 0 = no update, 1 = update.
> + * @acc_ae: 0 = no update, 1 = update.
> + * @acc_af: 0 = no update, 1 = update.
> + * @acc_awb: 0 = no update, 1 = update.
> + * @__acc_osys: 0 = no update, 1 = update.
> + * @reserved3: Not used.
> + * @lin_vmem_params: 0 = no update, 1 = update.
> + * @tnr3_vmem_params: 0 = no update, 1 = update.
> + * @xnr3_vmem_params: 0 = no update, 1 = update.
> + * @tnr3_dmem_params: 0 = no update, 1 = update.
> + * @xnr3_dmem_params: 0 = no update, 1 = update.
> + * @reserved4: Not used.
> + * @obgrid_param: 0 = no update, 1 = update.
> + * @reserved5: Not used.
> + */
> +struct ipu3_uapi_flags {
> + __u32 gdc:1;
> + __u32 obgrid:1;
> + __u32 reserved1:30;
> +
> + __u32 acc_bnr:1;
> + __u32 acc_green_disparity:1;
> + __u32 acc_dm:1;
> + __u32 acc_ccm:1;
> + __u32 acc_gamma:1;
> + __u32 acc_csc:1;
> + __u32 acc_cds:1;
> + __u32 acc_shd:1;
> + __u32 reserved2:2;
> + __u32 acc_iefd:1;
> + __u32 acc_yds_c0:1;
> + __u32 acc_chnr_c0:1;
> + __u32 acc_y_ee_nr:1;
> + __u32 acc_yds:1;
> + __u32 acc_chnr:1;
> + __u32 acc_ytm:1;
> + __u32 acc_yds2:1;
> + __u32 acc_tcc:1;
> + __u32 acc_dpc:1;
> + __u32 acc_bds:1;
> + __u32 acc_anr:1;
> + __u32 acc_awb_fr:1;
> + __u32 acc_ae:1;
> + __u32 acc_af:1;
> + __u32 acc_awb:1;
> + __u32 reserved3:4;
> +
> + __u32 lin_vmem_params:1;
> + __u32 tnr3_vmem_params:1;
> + __u32 xnr3_vmem_params:1;
> + __u32 tnr3_dmem_params:1;
> + __u32 xnr3_dmem_params:1;
> + __u32 reserved4:1;
> + __u32 obgrid_param:1;
> + __u32 reserved5:25;
> +} __attribute__((packed));
> +
> +/**
> + * struct ipu3_uapi_params - V4L2_META_FMT_IPU3_PARAMS
> + *
> + * @use: select which parameters to apply, see &ipu3_uapi_flags
> + * @acc_param: ACC parameters, as specified by &ipu3_uapi_acc_param
> + * @lin_vmem_params: linearization VMEM, as specified by
> + * &ipu3_uapi_isp_lin_vmem_params
> + * @tnr3_vmem_params: tnr3 VMEM as specified by
> + * &ipu3_uapi_isp_tnr3_vmem_params
> + * @xnr3_vmem_params: xnr3 VMEM as specified by
> + * &ipu3_uapi_isp_xnr3_vmem_params
> + * @tnr3_dmem_params: tnr3 DMEM as specified by &ipu3_uapi_isp_tnr3_params
> + * @xnr3_dmem_params: xnr3 DMEM as specified by &ipu3_uapi_isp_xnr3_params
> + * @obgrid_param: obgrid parameters as specified by
> + * &ipu3_uapi_obgrid_param
> + *
> + * The video queue "parameters" is of format V4L2_META_FMT_IPU3_PARAMS.
> + * This is a "single plane" v4l2_meta_format using V4L2_BUF_TYPE_META_OUTPUT.
> + *
> + * struct ipu3_uapi_params as defined below contains a lot of parameters and
> + * ipu3_uapi_flags selects which parameters to apply.
> + */
> +struct ipu3_uapi_params {
> + /* Flags which of the settings below are to be applied */
> + struct ipu3_uapi_flags use __attribute__((aligned(32)));
> +
> + /* Accelerator cluster parameters */
> + struct ipu3_uapi_acc_param acc_param;
> +
> + /* ISP vector address space parameters */
> + struct ipu3_uapi_isp_lin_vmem_params lin_vmem_params;
> + struct ipu3_uapi_isp_tnr3_vmem_params tnr3_vmem_params;
> + struct ipu3_uapi_isp_xnr3_vmem_params xnr3_vmem_params;
> +
> + /* ISP data memory (DMEM) parameters */
> + struct ipu3_uapi_isp_tnr3_params tnr3_dmem_params;
> + struct ipu3_uapi_isp_xnr3_params xnr3_dmem_params;
> +
> + /* Optical black level compensation */
> + struct ipu3_uapi_obgrid_param obgrid_param;
> +} __attribute__((packed));
> +
> +#endif /* __IPU3_UAPI_H */
> --
> 2.29.2
>
> _______________________________________________
> libcamera-devel mailing list
> libcamera-devel at lists.libcamera.org
> https://lists.libcamera.org/listinfo/libcamera-devel
More information about the libcamera-devel
mailing list