1/* SPDX-License-Identifier: LGPL-2.1+ */
2/*
3 * Copyright 2016 Tom aan de Wiel
4 * Copyright 2018 Cisco Systems, Inc. and/or its affiliates. All rights reserved.
5 */
6
7#ifndef CODEC_FWHT_H
8#define CODEC_FWHT_H
9
10#include <linux/types.h>
11#include <linux/bitops.h>
12#include <asm/byteorder.h>
13
14/*
15 * The compressed format consists of a fwht_cframe_hdr struct followed by the
16 * compressed frame data. The header contains the size of that data.
17 * Each Y, Cb and Cr plane is compressed separately. If the compressed
18 * size of each plane becomes larger than the uncompressed size, then
19 * that plane is stored uncompressed and the corresponding bit is set
20 * in the flags field of the header.
21 *
22 * Each compressed plane consists of macroblocks and each macroblock
23 * is run-length-encoded. Each macroblock starts with a 16 bit value.
24 * Bit 15 indicates if this is a P-coded macroblock (1) or not (0).
25 * P-coded macroblocks contain a delta against the previous frame.
26 *
27 * Bits 1-12 contain a number. If non-zero, then this same macroblock
28 * repeats that number of times. This results in a high degree of
29 * compression for generated images like colorbars.
30 *
31 * Following this macroblock header the MB coefficients are run-length
32 * encoded: the top 12 bits contain the coefficient, the bottom 4 bits
33 * tell how many times this coefficient occurs. The value 0xf indicates
34 * that the remainder of the macroblock should be filled with zeroes.
35 *
36 * All 16 and 32 bit values are stored in big-endian (network) order.
37 *
38 * Each fwht_cframe_hdr starts with an 8 byte magic header that is
39 * guaranteed not to occur in the compressed frame data. This header
40 * can be used to sync to the next frame.
41 *
42 * This codec uses the Fast Walsh Hadamard Transform. Tom aan de Wiel
43 * developed this as part of a university project, specifically for use
44 * with this driver. His project report can be found here:
45 *
46 * https://hverkuil.home.xs4all.nl/fwht.pdf
47 */
48
49/*
50 * This is a sequence of 8 bytes with the low 4 bits set to 0xf.
51 *
52 * This sequence cannot occur in the encoded data
53 *
54 * Note that these two magic values are symmetrical so endian issues here.
55 */
56#define FWHT_MAGIC1 0x4f4f4f4f
57#define FWHT_MAGIC2 0xffffffff
58
59#define FWHT_VERSION 3
60
61/* Set if this is an interlaced format */
62#define FWHT_FL_IS_INTERLACED		BIT(0)
63/* Set if this is a bottom-first (NTSC) interlaced format */
64#define FWHT_FL_IS_BOTTOM_FIRST		BIT(1)
65/* Set if each 'frame' contains just one field */
66#define FWHT_FL_IS_ALTERNATE		BIT(2)
67/*
68 * If FWHT_FL_IS_ALTERNATE was set, then this is set if this
69 * 'frame' is the bottom field, else it is the top field.
70 */
71#define FWHT_FL_IS_BOTTOM_FIELD		BIT(3)
72/* Set if this frame is uncompressed */
73#define FWHT_FL_LUMA_IS_UNCOMPRESSED	BIT(4)
74#define FWHT_FL_CB_IS_UNCOMPRESSED	BIT(5)
75#define FWHT_FL_CR_IS_UNCOMPRESSED	BIT(6)
76#define FWHT_FL_CHROMA_FULL_HEIGHT	BIT(7)
77#define FWHT_FL_CHROMA_FULL_WIDTH	BIT(8)
78#define FWHT_FL_ALPHA_IS_UNCOMPRESSED	BIT(9)
79#define FWHT_FL_I_FRAME			BIT(10)
80
81/* A 4-values flag - the number of components - 1 */
82#define FWHT_FL_COMPONENTS_NUM_MSK	GENMASK(18, 16)
83#define FWHT_FL_COMPONENTS_NUM_OFFSET	16
84
85#define FWHT_FL_PIXENC_MSK	GENMASK(20, 19)
86#define FWHT_FL_PIXENC_OFFSET	19
87#define FWHT_FL_PIXENC_YUV	(1 << FWHT_FL_PIXENC_OFFSET)
88#define FWHT_FL_PIXENC_RGB	(2 << FWHT_FL_PIXENC_OFFSET)
89#define FWHT_FL_PIXENC_HSV	(3 << FWHT_FL_PIXENC_OFFSET)
90
91/*
92 * A macro to calculate the needed padding in order to make sure
93 * both luma and chroma components resolutions are rounded up to
94 * a multiple of 8
95 */
96#define vic_round_dim(dim, div) (round_up((dim) / (div), 8) * (div))
97
98struct fwht_cframe_hdr {
99	u32 magic1;
100	u32 magic2;
101	__be32 version;
102	__be32 width, height;
103	__be32 flags;
104	__be32 colorspace;
105	__be32 xfer_func;
106	__be32 ycbcr_enc;
107	__be32 quantization;
108	__be32 size;
109};
110
111struct fwht_cframe {
112	u16 i_frame_qp;
113	u16 p_frame_qp;
114	__be16 *rlc_data;
115	s16 coeffs[8 * 8];
116	s16 de_coeffs[8 * 8];
117	s16 de_fwht[8 * 8];
118	u32 size;
119};
120
121struct fwht_raw_frame {
122	unsigned int width_div;
123	unsigned int height_div;
124	unsigned int luma_alpha_step;
125	unsigned int chroma_step;
126	unsigned int components_num;
127	u8 *buf;
128	u8 *luma, *cb, *cr, *alpha;
129};
130
131#define FWHT_FRAME_PCODED	BIT(0)
132#define FWHT_FRAME_UNENCODED	BIT(1)
133#define FWHT_LUMA_UNENCODED	BIT(2)
134#define FWHT_CB_UNENCODED	BIT(3)
135#define FWHT_CR_UNENCODED	BIT(4)
136#define FWHT_ALPHA_UNENCODED	BIT(5)
137
138u32 fwht_encode_frame(struct fwht_raw_frame *frm,
139		      struct fwht_raw_frame *ref_frm,
140		      struct fwht_cframe *cf,
141		      bool is_intra, bool next_is_intra,
142		      unsigned int width, unsigned int height,
143		      unsigned int stride, unsigned int chroma_stride);
144bool fwht_decode_frame(struct fwht_cframe *cf, u32 hdr_flags,
145		unsigned int components_num, unsigned int width,
146		unsigned int height, const struct fwht_raw_frame *ref,
147		unsigned int ref_stride, unsigned int ref_chroma_stride,
148		struct fwht_raw_frame *dst, unsigned int dst_stride,
149		unsigned int dst_chroma_stride);
150#endif
151