1/*
2 *  Fast C2P (Chunky-to-Planar) Conversion
3 *
4 *  Copyright (C) 2003 Geert Uytterhoeven
5 *
6 *  NOTES:
7 *    - This code was inspired by Scout's C2P tutorial
8 *    - It assumes to run on a big endian system
9 *
10 *  This file is subject to the terms and conditions of the GNU General Public
11 *  License. See the file COPYING in the main directory of this archive
12 *  for more details.
13 */
14
15#include <linux/string.h>
16#include "c2p.h"
17
18
19    /*
20     *  Basic transpose step
21     */
22
23#define _transp(d, i1, i2, shift, mask)			\
24    do {						\
25	u32 t = (d[i1] ^ (d[i2] >> shift)) & mask;	\
26	d[i1] ^= t;					\
27	d[i2] ^= t << shift;				\
28    } while (0)
29
30static inline u32 get_mask(int n)
31{
32    switch (n) {
33	case 1:
34	    return 0x55555555;
35	    break;
36
37	case 2:
38	    return 0x33333333;
39	    break;
40
41	case 4:
42	    return 0x0f0f0f0f;
43	    break;
44
45	case 8:
46	    return 0x00ff00ff;
47	    break;
48
49	case 16:
50	    return 0x0000ffff;
51	    break;
52    }
53    return 0;
54}
55
56#define transp_nx1(d, n)				\
57    do {						\
58	u32 mask = get_mask(n);				\
59	/* First block */				\
60	_transp(d, 0, 1, n, mask);			\
61	/* Second block */				\
62	_transp(d, 2, 3, n, mask);			\
63	/* Third block */				\
64	_transp(d, 4, 5, n, mask);			\
65	/* Fourth block */				\
66	_transp(d, 6, 7, n, mask);			\
67    } while (0)
68
69#define transp_nx2(d, n)				\
70    do {						\
71	u32 mask = get_mask(n);				\
72	/* First block */				\
73	_transp(d, 0, 2, n, mask);			\
74	_transp(d, 1, 3, n, mask);			\
75	/* Second block */				\
76	_transp(d, 4, 6, n, mask);			\
77	_transp(d, 5, 7, n, mask);			\
78    } while (0)
79
80#define transp_nx4(d, n)				\
81    do {						\
82	u32 mask = get_mask(n);				\
83	_transp(d, 0, 4, n, mask);			\
84	_transp(d, 1, 5, n, mask);			\
85	_transp(d, 2, 6, n, mask);			\
86	_transp(d, 3, 7, n, mask);			\
87    } while (0)
88
89#define transp(d, n, m)	transp_nx ## m(d, n)
90
91
92    /*
93     *  Perform a full C2P step on 32 8-bit pixels, stored in 8 32-bit words
94     *  containing
95     *    - 32 8-bit chunky pixels on input
96     *    - permuted planar data on output
97     */
98
99static void c2p_8bpp(u32 d[8])
100{
101    transp(d, 16, 4);
102    transp(d, 8, 2);
103    transp(d, 4, 1);
104    transp(d, 2, 4);
105    transp(d, 1, 2);
106}
107
108
109    /*
110     *  Array containing the permution indices of the planar data after c2p
111     */
112
113static const int perm_c2p_8bpp[8] = { 7, 5, 3, 1, 6, 4, 2, 0 };
114
115
116    /*
117     *  Compose two values, using a bitmask as decision value
118     *  This is equivalent to (a & mask) | (b & ~mask)
119     */
120
121static inline unsigned long comp(unsigned long a, unsigned long b,
122				 unsigned long mask)
123{
124	return ((a ^ b) & mask) ^ b;
125}
126
127
128    /*
129     *  Store a full block of planar data after c2p conversion
130     */
131
132static inline void store_planar(char *dst, u32 dst_inc, u32 bpp, u32 d[8])
133{
134    int i;
135
136    for (i = 0; i < bpp; i++, dst += dst_inc)
137	*(u32 *)dst = d[perm_c2p_8bpp[i]];
138}
139
140
141    /*
142     *  Store a partial block of planar data after c2p conversion
143     */
144
145static inline void store_planar_masked(char *dst, u32 dst_inc, u32 bpp,
146				       u32 d[8], u32 mask)
147{
148    int i;
149
150    for (i = 0; i < bpp; i++, dst += dst_inc)
151	*(u32 *)dst = comp(d[perm_c2p_8bpp[i]], *(u32 *)dst, mask);
152}
153
154
155    /*
156     *  c2p - Copy 8-bit chunky image data to a planar frame buffer
157     *  @dst: Starting address of the planar frame buffer
158     *  @dx: Horizontal destination offset (in pixels)
159     *  @dy: Vertical destination offset (in pixels)
160     *  @width: Image width (in pixels)
161     *  @height: Image height (in pixels)
162     *  @dst_nextline: Frame buffer offset to the next line (in bytes)
163     *  @dst_nextplane: Frame buffer offset to the next plane (in bytes)
164     *  @src_nextline: Image offset to the next line (in bytes)
165     *  @bpp: Bits per pixel of the planar frame buffer (1-8)
166     */
167
168void c2p(u8 *dst, const u8 *src, u32 dx, u32 dy, u32 width, u32 height,
169	 u32 dst_nextline, u32 dst_nextplane, u32 src_nextline, u32 bpp)
170{
171    int dst_idx;
172    u32 d[8], first, last, w;
173    const u8 *c;
174    u8 *p;
175
176    dst += dy*dst_nextline+(dx & ~31);
177    dst_idx = dx % 32;
178    first = ~0UL >> dst_idx;
179    last = ~(~0UL >> ((dst_idx+width) % 32));
180    while (height--) {
181	c = src;
182	p = dst;
183	w = width;
184	if (dst_idx+width <= 32) {
185	    /* Single destination word */
186	    first &= last;
187	    memset(d, 0, sizeof(d));
188	    memcpy((u8 *)d+dst_idx, c, width);
189	    c += width;
190	    c2p_8bpp(d);
191	    store_planar_masked(p, dst_nextplane, bpp, d, first);
192	    p += 4;
193	} else {
194	    /* Multiple destination words */
195	    w = width;
196	    /* Leading bits */
197	    if (dst_idx) {
198		w = 32 - dst_idx;
199		memset(d, 0, dst_idx);
200		memcpy((u8 *)d+dst_idx, c, w);
201		c += w;
202		c2p_8bpp(d);
203		store_planar_masked(p, dst_nextplane, bpp, d, first);
204		p += 4;
205		w = width-w;
206	    }
207	    /* Main chunk */
208	    while (w >= 32) {
209		memcpy(d, c, 32);
210		c += 32;
211		c2p_8bpp(d);
212		store_planar(p, dst_nextplane, bpp, d);
213		p += 4;
214		w -= 32;
215	    }
216	    /* Trailing bits */
217	    w %= 32;
218	    if (w > 0) {
219		memcpy(d, c, w);
220		memset((u8 *)d+w, 0, 32-w);
221		c2p_8bpp(d);
222		store_planar_masked(p, dst_nextplane, bpp, d, last);
223	    }
224	}
225	src += src_nextline;
226	dst += dst_nextline;
227    }
228}
229