drivers/video/c2p.c

   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/module.h>
  16 #include <linux/string.h>
  17 #include "c2p.h"
  18
  19
  20     /*
  21      *  Basic transpose step
  22      */
  23
  24 #define _transp(d, i1, i2, shift, mask)                 \
  25     do {                                                \
  26         u32 t = (d[i1] ^ (d[i2] >> shift)) & mask;      \
  27         d[i1] ^= t;                                     \
  28         d[i2] ^= t << shift;                            \
  29     } while (0)
  30
  31 static inline u32 get_mask(int n)
  32 {
  33     switch (n) {
  34         case 1:
  35             return 0x55555555;
  36             break;
  37
  38         case 2:
  39             return 0x33333333;
  40             break;
  41
  42         case 4:
  43             return 0x0f0f0f0f;
  44             break;
  45
  46         case 8:
  47             return 0x00ff00ff;
  48             break;
  49
  50         case 16:
  51             return 0x0000ffff;
  52             break;
  53     }
  54     return 0;
  55 }
  56
  57 #define transp_nx1(d, n)                                \
  58     do {                                                \
  59         u32 mask = get_mask(n);                         \
  60         /* First block */                               \
  61         _transp(d, 0, 1, n, mask);                      \
  62         /* Second block */                              \
  63         _transp(d, 2, 3, n, mask);                      \
  64         /* Third block */                               \
  65         _transp(d, 4, 5, n, mask);                      \
  66         /* Fourth block */                              \
  67         _transp(d, 6, 7, n, mask);                      \
  68     } while (0)
  69
  70 #define transp_nx2(d, n)                                \
  71     do {                                                \
  72         u32 mask = get_mask(n);                         \
  73         /* First block */                               \
  74         _transp(d, 0, 2, n, mask);                      \
  75         _transp(d, 1, 3, n, mask);                      \
  76         /* Second block */                              \
  77         _transp(d, 4, 6, n, mask);                      \
  78         _transp(d, 5, 7, n, mask);                      \
  79     } while (0)
  80
  81 #define transp_nx4(d, n)                                \
  82     do {                                                \
  83         u32 mask = get_mask(n);                         \
  84         _transp(d, 0, 4, n, mask);                      \
  85         _transp(d, 1, 5, n, mask);                      \
  86         _transp(d, 2, 6, n, mask);                      \
  87         _transp(d, 3, 7, n, mask);                      \
  88     } while (0)
  89
  90 #define transp(d, n, m) transp_nx ## m(d, n)
  91
  92
  93     /*
  94      *  Perform a full C2P step on 32 8-bit pixels, stored in 8 32-bit words
  95      *  containing
  96      *    - 32 8-bit chunky pixels on input
  97      *    - permuted planar data on output
  98      */
  99
 100 static void c2p_8bpp(u32 d[8])
 101 {
 102     transp(d, 16, 4);
 103     transp(d, 8, 2);
 104     transp(d, 4, 1);
 105     transp(d, 2, 4);
 106     transp(d, 1, 2);
 107 }
 108
 109
 110     /*
 111      *  Array containing the permution indices of the planar data after c2p
 112      */
 113
 114 static const int perm_c2p_8bpp[8] = { 7, 5, 3, 1, 6, 4, 2, 0 };
 115
 116
 117     /*
 118      *  Compose two values, using a bitmask as decision value
 119      *  This is equivalent to (a & mask) | (b & ~mask)
 120      */
 121
 122 static inline unsigned long comp(unsigned long a, unsigned long b,
 123                                  unsigned long mask)
 124 {
 125         return ((a ^ b) & mask) ^ b;
 126 }
 127
 128
 129     /*
 130      *  Store a full block of planar data after c2p conversion
 131      */
 132
 133 static inline void store_planar(char *dst, u32 dst_inc, u32 bpp, u32 d[8])
 134 {
 135     int i;
 136
 137     for (i = 0; i < bpp; i++, dst += dst_inc)
 138         *(u32 *)dst = d[perm_c2p_8bpp[i]];
 139 }
 140
 141
 142     /*
 143      *  Store a partial block of planar data after c2p conversion
 144      */
 145
 146 static inline void store_planar_masked(char *dst, u32 dst_inc, u32 bpp,
 147                                        u32 d[8], u32 mask)
 148 {
 149     int i;
 150
 151     for (i = 0; i < bpp; i++, dst += dst_inc)
 152         *(u32 *)dst = comp(d[perm_c2p_8bpp[i]], *(u32 *)dst, mask);
 153 }
 154
 155
 156     /*
 157      *  c2p - Copy 8-bit chunky image data to a planar frame buffer
 158      *  @dst: Starting address of the planar frame buffer
 159      *  @dx: Horizontal destination offset (in pixels)
 160      *  @dy: Vertical destination offset (in pixels)
 161      *  @width: Image width (in pixels)
 162      *  @height: Image height (in pixels)
 163      *  @dst_nextline: Frame buffer offset to the next line (in bytes)
 164      *  @dst_nextplane: Frame buffer offset to the next plane (in bytes)
 165      *  @src_nextline: Image offset to the next line (in bytes)
 166      *  @bpp: Bits per pixel of the planar frame buffer (1-8)
 167      */
 168
 169 void c2p(u8 *dst, const u8 *src, u32 dx, u32 dy, u32 width, u32 height,
 170          u32 dst_nextline, u32 dst_nextplane, u32 src_nextline, u32 bpp)
 171 {
 172     int dst_idx;
 173     u32 d[8], first, last, w;
 174     const u8 *c;
 175     u8 *p;
 176
 177     dst += dy*dst_nextline+(dx & ~31);
 178     dst_idx = dx % 32;
 179     first = ~0UL >> dst_idx;
 180     last = ~(~0UL >> ((dst_idx+width) % 32));
 181     while (height--) {
 182         c = src;
 183         p = dst;
 184         w = width;
 185         if (dst_idx+width <= 32) {
 186             /* Single destination word */
 187             first &= last;
 188             memset(d, 0, sizeof(d));
 189             memcpy((u8 *)d+dst_idx, c, width);
 190             c += width;
 191             c2p_8bpp(d);
 192             store_planar_masked(p, dst_nextplane, bpp, d, first);
 193             p += 4;
 194         } else {
 195             /* Multiple destination words */
 196             w = width;
 197             /* Leading bits */
 198             if (dst_idx) {
 199                 w = 32 - dst_idx;
 200                 memset(d, 0, dst_idx);
 201                 memcpy((u8 *)d+dst_idx, c, w);
 202                 c += w;
 203                 c2p_8bpp(d);
 204                 store_planar_masked(p, dst_nextplane, bpp, d, first);
 205                 p += 4;
 206                 w = width-w;
 207             }
 208             /* Main chunk */
 209             while (w >= 32) {
 210                 memcpy(d, c, 32);
 211                 c += 32;
 212                 c2p_8bpp(d);
 213                 store_planar(p, dst_nextplane, bpp, d);
 214                 p += 4;
 215                 w -= 32;
 216             }
 217             /* Trailing bits */
 218             w %= 32;
 219             if (w > 0) {
 220                 memcpy(d, c, w);
 221                 memset((u8 *)d+w, 0, 32-w);
 222                 c2p_8bpp(d);
 223                 store_planar_masked(p, dst_nextplane, bpp, d, last);
 224             }
 225         }
 226         src += src_nextline;
 227         dst += dst_nextline;
 228     }
 229 }
 230 EXPORT_SYMBOL_GPL(c2p);
 231
 232 MODULE_LICENSE("GPL");