Details | Last modification | View Log | RSS feed
| Rev | Author | Line No. | Line |
|---|---|---|---|
| 468 | giacomo | 1 | /* |
| 2 | * Generic function for frame buffer with packed pixels of any depth. |
||
| 3 | * |
||
| 4 | * Copyright (C) June 1999 James Simmons |
||
| 5 | * |
||
| 6 | * This file is subject to the terms and conditions of the GNU General Public |
||
| 7 | * License. See the file COPYING in the main directory of this archive for |
||
| 8 | * more details. |
||
| 9 | * |
||
| 10 | * NOTES: |
||
| 11 | * |
||
| 12 | * This is for cfb packed pixels. Iplan and such are incorporated in the |
||
| 13 | * drivers that need them. |
||
| 14 | * |
||
| 15 | * FIXME |
||
| 16 | * The code for 24 bit is horrible. It copies byte by byte size instead of |
||
| 17 | * longs like the other sizes. Needs to be optimized. |
||
| 18 | * |
||
| 19 | * Also need to add code to deal with cards endians that are different than |
||
| 20 | * the native cpu endians. I also need to deal with MSB position in the word. |
||
| 21 | * |
||
| 22 | */ |
||
| 23 | |||
| 24 | #include <linuxcomp.h> |
||
| 25 | |||
| 26 | #include <linux/config.h> |
||
| 27 | #include <linux/module.h> |
||
| 28 | #include <linux/kernel.h> |
||
| 29 | #include <linux/string.h> |
||
| 30 | #include <linux/fb.h> |
||
| 31 | #include <linux/slab.h> |
||
| 32 | #include <asm/types.h> |
||
| 33 | #include <asm/io.h> |
||
| 34 | |||
| 35 | #define LONG_MASK (BITS_PER_LONG - 1) |
||
| 36 | |||
| 37 | #if BITS_PER_LONG == 32 |
||
| 38 | #define FB_WRITEL fb_writel |
||
| 39 | #define FB_READL fb_readl |
||
| 40 | #define SHIFT_PER_LONG 5 |
||
| 41 | #define BYTES_PER_LONG 4 |
||
| 42 | #else |
||
| 43 | #define FB_WRITEL fb_writeq |
||
| 44 | #define FB_READL fb_readq |
||
| 45 | #define SHIFT_PER_LONG 6 |
||
| 46 | #define BYTES_PER_LONG 8 |
||
| 47 | #endif |
||
| 48 | |||
| 49 | static void bitcpy(unsigned long *dst, int dst_idx, const unsigned long *src, |
||
| 50 | int src_idx, unsigned long n) |
||
| 51 | { |
||
| 52 | unsigned long first, last; |
||
| 53 | int shift = dst_idx-src_idx, left, right; |
||
| 54 | unsigned long d0, d1; |
||
| 55 | int m; |
||
| 56 | |||
| 57 | if (!n) |
||
| 58 | return; |
||
| 59 | |||
| 60 | shift = dst_idx-src_idx; |
||
| 61 | first = ~0UL >> dst_idx; |
||
| 62 | last = ~(~0UL >> ((dst_idx+n) % BITS_PER_LONG)); |
||
| 63 | |||
| 64 | if (!shift) { |
||
| 65 | // Same alignment for source and dest |
||
| 66 | |||
| 67 | if (dst_idx+n <= BITS_PER_LONG) { |
||
| 68 | // Single word |
||
| 69 | if (last) |
||
| 70 | first &= last; |
||
| 71 | FB_WRITEL((FB_READL(src) & first) | (FB_READL(dst) & ~first), dst); |
||
| 72 | } else { |
||
| 73 | // Multiple destination words |
||
| 74 | // Leading bits |
||
| 75 | if (first) { |
||
| 76 | |||
| 77 | FB_WRITEL((FB_READL(src) & first) | |
||
| 78 | (FB_READL(dst) & ~first), dst); |
||
| 79 | dst++; |
||
| 80 | src++; |
||
| 81 | n -= BITS_PER_LONG-dst_idx; |
||
| 82 | } |
||
| 83 | |||
| 84 | // Main chunk |
||
| 85 | n /= BITS_PER_LONG; |
||
| 86 | while (n >= 8) { |
||
| 87 | FB_WRITEL(FB_READL(src++), dst++); |
||
| 88 | FB_WRITEL(FB_READL(src++), dst++); |
||
| 89 | FB_WRITEL(FB_READL(src++), dst++); |
||
| 90 | FB_WRITEL(FB_READL(src++), dst++); |
||
| 91 | FB_WRITEL(FB_READL(src++), dst++); |
||
| 92 | FB_WRITEL(FB_READL(src++), dst++); |
||
| 93 | FB_WRITEL(FB_READL(src++), dst++); |
||
| 94 | FB_WRITEL(FB_READL(src++), dst++); |
||
| 95 | n -= 8; |
||
| 96 | } |
||
| 97 | while (n--) |
||
| 98 | FB_WRITEL(FB_READL(src++), dst++); |
||
| 99 | // Trailing bits |
||
| 100 | if (last) |
||
| 101 | FB_WRITEL((FB_READL(src) & last) | (FB_READL(dst) & ~last), dst); |
||
| 102 | } |
||
| 103 | } else { |
||
| 104 | // Different alignment for source and dest |
||
| 105 | |||
| 106 | right = shift & (BITS_PER_LONG-1); |
||
| 107 | left = -shift & (BITS_PER_LONG-1); |
||
| 108 | |||
| 109 | if (dst_idx+n <= BITS_PER_LONG) { |
||
| 110 | // Single destination word |
||
| 111 | if (last) |
||
| 112 | first &= last; |
||
| 113 | if (shift > 0) { |
||
| 114 | // Single source word |
||
| 115 | FB_WRITEL(((FB_READL(src) >> right) & first) | |
||
| 116 | (FB_READL(dst) & ~first), dst); |
||
| 117 | } else if (src_idx+n <= BITS_PER_LONG) { |
||
| 118 | // Single source word |
||
| 119 | FB_WRITEL(((FB_READL(src) << left) & first) | |
||
| 120 | (FB_READL(dst) & ~first), dst); |
||
| 121 | } else { |
||
| 122 | // 2 source words |
||
| 123 | d0 = FB_READL(src++); |
||
| 124 | d1 = FB_READL(src); |
||
| 125 | FB_WRITEL(((d0<<left | d1>>right) & first) | |
||
| 126 | (FB_READL(dst) & ~first), dst); |
||
| 127 | } |
||
| 128 | } else { |
||
| 129 | // Multiple destination words |
||
| 130 | d0 = FB_READL(src++); |
||
| 131 | // Leading bits |
||
| 132 | if (shift > 0) { |
||
| 133 | // Single source word |
||
| 134 | FB_WRITEL(((d0 >> right) & first) | |
||
| 135 | (FB_READL(dst) & ~first), dst); |
||
| 136 | dst++; |
||
| 137 | n -= BITS_PER_LONG-dst_idx; |
||
| 138 | } else { |
||
| 139 | // 2 source words |
||
| 140 | d1 = FB_READL(src++); |
||
| 141 | FB_WRITEL(((d0<<left | d1>>right) & first) | |
||
| 142 | (FB_READL(dst) & ~first), dst); |
||
| 143 | d0 = d1; |
||
| 144 | dst++; |
||
| 145 | n -= BITS_PER_LONG-dst_idx; |
||
| 146 | } |
||
| 147 | |||
| 148 | // Main chunk |
||
| 149 | m = n % BITS_PER_LONG; |
||
| 150 | n /= BITS_PER_LONG; |
||
| 151 | while (n >= 4) { |
||
| 152 | d1 = FB_READL(src++); |
||
| 153 | FB_WRITEL(d0 << left | d1 >> right, dst++); |
||
| 154 | d0 = d1; |
||
| 155 | d1 = FB_READL(src++); |
||
| 156 | FB_WRITEL(d0 << left | d1 >> right, dst++); |
||
| 157 | d0 = d1; |
||
| 158 | d1 = FB_READL(src++); |
||
| 159 | FB_WRITEL(d0 << left | d1 >> right, dst++); |
||
| 160 | d0 = d1; |
||
| 161 | d1 = FB_READL(src++); |
||
| 162 | FB_WRITEL(d0 << left | d1 >> right, dst++); |
||
| 163 | d0 = d1; |
||
| 164 | n -= 4; |
||
| 165 | } |
||
| 166 | while (n--) { |
||
| 167 | d1 = FB_READL(src++); |
||
| 168 | FB_WRITEL(d0 << left | d1 >> right, dst++); |
||
| 169 | d0 = d1; |
||
| 170 | } |
||
| 171 | |||
| 172 | // Trailing bits |
||
| 173 | if (last) { |
||
| 174 | if (m <= right) { |
||
| 175 | // Single source word |
||
| 176 | FB_WRITEL(((d0 << left) & last) | |
||
| 177 | (FB_READL(dst) & ~last), |
||
| 178 | dst); |
||
| 179 | } else { |
||
| 180 | // 2 source words |
||
| 181 | d1 = FB_READL(src); |
||
| 182 | FB_WRITEL(((d0<<left | d1>>right) & |
||
| 183 | last) | (FB_READL(dst) & |
||
| 184 | ~last), dst); |
||
| 185 | } |
||
| 186 | } |
||
| 187 | } |
||
| 188 | } |
||
| 189 | } |
||
| 190 | |||
| 191 | static void bitcpy_rev(unsigned long *dst, int dst_idx, |
||
| 192 | const unsigned long *src, int src_idx, unsigned long n) |
||
| 193 | { |
||
| 194 | unsigned long first, last; |
||
| 195 | int shift = dst_idx-src_idx, left, right; |
||
| 196 | unsigned long d0, d1; |
||
| 197 | int m; |
||
| 198 | |||
| 199 | if (!n) |
||
| 200 | return; |
||
| 201 | |||
| 202 | dst += (n-1)/BITS_PER_LONG; |
||
| 203 | src += (n-1)/BITS_PER_LONG; |
||
| 204 | if ((n-1) % BITS_PER_LONG) { |
||
| 205 | dst_idx += (n-1) % BITS_PER_LONG; |
||
| 206 | dst += dst_idx >> SHIFT_PER_LONG; |
||
| 207 | dst_idx &= BITS_PER_LONG-1; |
||
| 208 | src_idx += (n-1) % BITS_PER_LONG; |
||
| 209 | src += src_idx >> SHIFT_PER_LONG; |
||
| 210 | src_idx &= BITS_PER_LONG-1; |
||
| 211 | } |
||
| 212 | |||
| 213 | shift = dst_idx-src_idx; |
||
| 214 | first = ~0UL << (BITS_PER_LONG-1-dst_idx); |
||
| 215 | last = ~(~0UL << (BITS_PER_LONG-1-((dst_idx-n) % BITS_PER_LONG))); |
||
| 216 | |||
| 217 | if (!shift) { |
||
| 218 | // Same alignment for source and dest |
||
| 219 | |||
| 220 | if ((unsigned long)dst_idx+1 >= n) { |
||
| 221 | // Single word |
||
| 222 | if (last) |
||
| 223 | first &= last; |
||
| 224 | FB_WRITEL((FB_READL(src) & first) | (FB_READL(dst) & ~first), dst); |
||
| 225 | } else { |
||
| 226 | // Multiple destination words |
||
| 227 | // Leading bits |
||
| 228 | if (first) { |
||
| 229 | FB_WRITEL((FB_READL(src) & first) | (FB_READL(dst) & ~first), dst); |
||
| 230 | dst--; |
||
| 231 | src--; |
||
| 232 | n -= dst_idx+1; |
||
| 233 | } |
||
| 234 | |||
| 235 | // Main chunk |
||
| 236 | n /= BITS_PER_LONG; |
||
| 237 | while (n >= 8) { |
||
| 238 | FB_WRITEL(FB_READL(src--), dst--); |
||
| 239 | FB_WRITEL(FB_READL(src--), dst--); |
||
| 240 | FB_WRITEL(FB_READL(src--), dst--); |
||
| 241 | FB_WRITEL(FB_READL(src--), dst--); |
||
| 242 | FB_WRITEL(FB_READL(src--), dst--); |
||
| 243 | FB_WRITEL(FB_READL(src--), dst--); |
||
| 244 | FB_WRITEL(FB_READL(src--), dst--); |
||
| 245 | FB_WRITEL(FB_READL(src--), dst--); |
||
| 246 | n -= 8; |
||
| 247 | } |
||
| 248 | while (n--) |
||
| 249 | FB_WRITEL(FB_READL(src--), dst--); |
||
| 250 | |||
| 251 | // Trailing bits |
||
| 252 | if (last) |
||
| 253 | FB_WRITEL((FB_READL(src) & last) | (FB_READL(dst) & ~last), dst); |
||
| 254 | } |
||
| 255 | } else { |
||
| 256 | // Different alignment for source and dest |
||
| 257 | |||
| 258 | right = shift & (BITS_PER_LONG-1); |
||
| 259 | left = -shift & (BITS_PER_LONG-1); |
||
| 260 | |||
| 261 | if ((unsigned long)dst_idx+1 >= n) { |
||
| 262 | // Single destination word |
||
| 263 | if (last) |
||
| 264 | first &= last; |
||
| 265 | if (shift < 0) { |
||
| 266 | // Single source word |
||
| 267 | FB_WRITEL((FB_READL(src) << left & first) | |
||
| 268 | (FB_READL(dst) & ~first), dst); |
||
| 269 | } else if (1+(unsigned long)src_idx >= n) { |
||
| 270 | // Single source word |
||
| 271 | FB_WRITEL(((FB_READL(src) >> right) & first) | |
||
| 272 | (FB_READL(dst) & ~first), dst); |
||
| 273 | } else { |
||
| 274 | // 2 source words |
||
| 275 | d0 = FB_READL(src--); |
||
| 276 | d1 = FB_READL(src); |
||
| 277 | FB_WRITEL(((d0>>right | d1<<left) & first) | |
||
| 278 | (FB_READL(dst) & ~first), dst); |
||
| 279 | } |
||
| 280 | } else { |
||
| 281 | // Multiple destination words |
||
| 282 | d0 = FB_READL(src--); |
||
| 283 | // Leading bits |
||
| 284 | if (shift < 0) { |
||
| 285 | // Single source word |
||
| 286 | FB_WRITEL(((d0 << left) & first) | |
||
| 287 | (FB_READL(dst) & ~first), dst); |
||
| 288 | dst--; |
||
| 289 | n -= dst_idx+1; |
||
| 290 | } else { |
||
| 291 | // 2 source words |
||
| 292 | d1 = FB_READL(src--); |
||
| 293 | FB_WRITEL(((d0>>right | d1<<left) & first) | |
||
| 294 | (FB_READL(dst) & ~first), dst); |
||
| 295 | d0 = d1; |
||
| 296 | dst--; |
||
| 297 | n -= dst_idx+1; |
||
| 298 | } |
||
| 299 | |||
| 300 | // Main chunk |
||
| 301 | m = n % BITS_PER_LONG; |
||
| 302 | n /= BITS_PER_LONG; |
||
| 303 | while (n >= 4) { |
||
| 304 | d1 = FB_READL(src--); |
||
| 305 | FB_WRITEL(d0 >> right | d1 << left, dst--); |
||
| 306 | d0 = d1; |
||
| 307 | d1 = FB_READL(src--); |
||
| 308 | FB_WRITEL(d0 >> right | d1 << left, dst--); |
||
| 309 | d0 = d1; |
||
| 310 | d1 = FB_READL(src--); |
||
| 311 | FB_WRITEL(d0 >> right | d1 << left, dst--); |
||
| 312 | d0 = d1; |
||
| 313 | d1 = FB_READL(src--); |
||
| 314 | FB_WRITEL(d0 >> right | d1 << left, dst--); |
||
| 315 | d0 = d1; |
||
| 316 | n -= 4; |
||
| 317 | } |
||
| 318 | while (n--) { |
||
| 319 | d1 = FB_READL(src--); |
||
| 320 | FB_WRITEL(d0 >> right | d1 << left, dst--); |
||
| 321 | d0 = d1; |
||
| 322 | } |
||
| 323 | |||
| 324 | // Trailing bits |
||
| 325 | if (last) { |
||
| 326 | if (m <= left) { |
||
| 327 | // Single source word |
||
| 328 | FB_WRITEL(((d0 >> right) & last) | |
||
| 329 | (FB_READL(dst) & ~last), |
||
| 330 | dst); |
||
| 331 | } else { |
||
| 332 | // 2 source words |
||
| 333 | d1 = FB_READL(src); |
||
| 334 | FB_WRITEL(((d0>>right | d1<<left) & |
||
| 335 | last) | (FB_READL(dst) & |
||
| 336 | ~last), dst); |
||
| 337 | } |
||
| 338 | } |
||
| 339 | } |
||
| 340 | } |
||
| 341 | } |
||
| 342 | |||
| 343 | void cfb_copyarea(struct fb_info *p, const struct fb_copyarea *area) |
||
| 344 | { |
||
| 345 | u32 dx = area->dx, dy = area->dy, sx = area->sx, sy = area->sy; |
||
| 346 | u32 height = area->height, width = area->width; |
||
| 347 | int x2, y2, old_dx, old_dy, vxres, vyres; |
||
| 348 | unsigned long next_line = p->fix.line_length; |
||
| 349 | int dst_idx = 0, src_idx = 0, rev_copy = 0; |
||
| 350 | unsigned long *dst = NULL, *src = NULL; |
||
| 351 | |||
| 352 | /* We want rotation but lack hardware to do it for us. */ |
||
| 353 | if (!p->fbops->fb_rotate && p->var.rotate) { |
||
| 354 | } |
||
| 355 | |||
| 356 | vxres = p->var.xres_virtual; |
||
| 357 | vyres = p->var.yres_virtual; |
||
| 358 | |||
| 359 | if (area->dx > vxres || area->sx > vxres || |
||
| 360 | area->dy > vyres || area->sy > vyres) |
||
| 361 | return; |
||
| 362 | |||
| 363 | /* clip the destination */ |
||
| 364 | old_dx = area->dx; |
||
| 365 | old_dy = area->dy; |
||
| 366 | |||
| 367 | /* |
||
| 368 | * We could use hardware clipping but on many cards you get around |
||
| 369 | * hardware clipping by writing to framebuffer directly. |
||
| 370 | */ |
||
| 371 | x2 = area->dx + area->width; |
||
| 372 | y2 = area->dy + area->height; |
||
| 373 | dx = area->dx > 0 ? area->dx : 0; |
||
| 374 | dy = area->dy > 0 ? area->dy : 0; |
||
| 375 | x2 = x2 < vxres ? x2 : vxres; |
||
| 376 | y2 = y2 < vyres ? y2 : vyres; |
||
| 377 | width = x2 - dx; |
||
| 378 | height = y2 - dy; |
||
| 379 | |||
| 380 | /* update sx1,sy1 */ |
||
| 381 | sx += (dx - old_dx); |
||
| 382 | sy += (dy - old_dy); |
||
| 383 | |||
| 384 | /* the source must be completely inside the virtual screen */ |
||
| 385 | if (sx < 0 || sy < 0 || |
||
| 386 | (sx + width) > vxres || |
||
| 387 | (sy + height) > vyres) |
||
| 388 | return; |
||
| 389 | |||
| 390 | if ((dy == sy && dx > sx) || |
||
| 391 | (dy > sy)) { |
||
| 392 | dy += height; |
||
| 393 | sy += height; |
||
| 394 | rev_copy = 1; |
||
| 395 | } |
||
| 396 | |||
| 397 | dst = src = (unsigned long *)((unsigned long)p->screen_base & |
||
| 398 | ~(BYTES_PER_LONG-1)); |
||
| 399 | dst_idx = src_idx = (unsigned long)p->screen_base & (BYTES_PER_LONG-1); |
||
| 400 | dst_idx += dy*next_line*8 + dx*p->var.bits_per_pixel; |
||
| 401 | src_idx += sy*next_line*8 + sx*p->var.bits_per_pixel; |
||
| 402 | |||
| 403 | if (p->fbops->fb_sync) |
||
| 404 | p->fbops->fb_sync(p); |
||
| 405 | if (rev_copy) { |
||
| 406 | while (height--) { |
||
| 407 | dst_idx -= next_line*8; |
||
| 408 | src_idx -= next_line*8; |
||
| 409 | dst += dst_idx >> SHIFT_PER_LONG; |
||
| 410 | dst_idx &= (BYTES_PER_LONG-1); |
||
| 411 | src += src_idx >> SHIFT_PER_LONG; |
||
| 412 | src_idx &= (BYTES_PER_LONG-1); |
||
| 413 | bitcpy_rev(dst, dst_idx, src, src_idx, |
||
| 414 | width*p->var.bits_per_pixel); |
||
| 415 | } |
||
| 416 | } else { |
||
| 417 | while (height--) { |
||
| 418 | dst += dst_idx >> SHIFT_PER_LONG; |
||
| 419 | dst_idx &= (BYTES_PER_LONG-1); |
||
| 420 | src += src_idx >> SHIFT_PER_LONG; |
||
| 421 | src_idx &= (BYTES_PER_LONG-1); |
||
| 422 | bitcpy(dst, dst_idx, src, src_idx, |
||
| 423 | width*p->var.bits_per_pixel); |
||
| 424 | dst_idx += next_line*8; |
||
| 425 | src_idx += next_line*8; |
||
| 426 | } |
||
| 427 | } |
||
| 428 | } |
||
| 429 | |||
| 430 | EXPORT_SYMBOL(cfb_copyarea); |
||
| 431 | |||
| 432 | MODULE_AUTHOR("James Simmons <jsimmons@users.sf.net>"); |
||
| 433 | MODULE_DESCRIPTION("Generic software accelerated copyarea"); |
||
| 434 | MODULE_LICENSE("GPL"); |
||
| 435 |