Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * Generic function for frame buffer with packed pixels of any depth. | |
3 | * | |
4 | * Copyright (C) 1999-2005 James Simmons <jsimmons@www.infradead.org> | |
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 | * | |
17 | * Also need to add code to deal with cards endians that are different than | |
18 | * the native cpu endians. I also need to deal with MSB position in the word. | |
19 | * | |
20 | * The two functions or copying forward and backward could be split up like | |
21 | * the ones for filling, i.e. in aligned and unaligned versions. This would | |
22 | * help moving some redundant computations and branches out of the loop, too. | |
23 | */ | |
24 | ||
1da177e4 LT |
25 | #include <linux/module.h> |
26 | #include <linux/kernel.h> | |
27 | #include <linux/string.h> | |
28 | #include <linux/fb.h> | |
29 | #include <linux/slab.h> | |
30 | #include <asm/types.h> | |
31 | #include <asm/io.h> | |
dc0e6e05 | 32 | #include "fb_draw.h" |
1da177e4 LT |
33 | |
34 | #if BITS_PER_LONG == 32 | |
35 | # define FB_WRITEL fb_writel | |
36 | # define FB_READL fb_readl | |
37 | #else | |
38 | # define FB_WRITEL fb_writeq | |
39 | # define FB_READL fb_readq | |
40 | #endif | |
41 | ||
1da177e4 LT |
42 | /* |
43 | * Generic bitwise copy algorithm | |
44 | */ | |
45 | ||
46 | static void | |
47 | bitcpy(unsigned long __iomem *dst, int dst_idx, const unsigned long __iomem *src, | |
48 | int src_idx, int bits, unsigned n) | |
49 | { | |
50 | unsigned long first, last; | |
51 | int const shift = dst_idx-src_idx; | |
52 | int left, right; | |
53 | ||
be0d9b6c AD |
54 | first = FB_SHIFT_HIGH(~0UL, dst_idx); |
55 | last = ~(FB_SHIFT_HIGH(~0UL, (dst_idx+n) % bits)); | |
1da177e4 LT |
56 | |
57 | if (!shift) { | |
58 | // Same alignment for source and dest | |
59 | ||
60 | if (dst_idx+n <= bits) { | |
61 | // Single word | |
62 | if (last) | |
63 | first &= last; | |
64 | FB_WRITEL( comp( FB_READL(src), FB_READL(dst), first), dst); | |
65 | } else { | |
66 | // Multiple destination words | |
67 | ||
68 | // Leading bits | |
69 | if (first != ~0UL) { | |
70 | FB_WRITEL( comp( FB_READL(src), FB_READL(dst), first), dst); | |
71 | dst++; | |
72 | src++; | |
73 | n -= bits - dst_idx; | |
74 | } | |
75 | ||
76 | // Main chunk | |
77 | n /= bits; | |
78 | while (n >= 8) { | |
79 | FB_WRITEL(FB_READL(src++), dst++); | |
80 | FB_WRITEL(FB_READL(src++), dst++); | |
81 | FB_WRITEL(FB_READL(src++), dst++); | |
82 | FB_WRITEL(FB_READL(src++), dst++); | |
83 | FB_WRITEL(FB_READL(src++), dst++); | |
84 | FB_WRITEL(FB_READL(src++), dst++); | |
85 | FB_WRITEL(FB_READL(src++), dst++); | |
86 | FB_WRITEL(FB_READL(src++), dst++); | |
87 | n -= 8; | |
88 | } | |
89 | while (n--) | |
90 | FB_WRITEL(FB_READL(src++), dst++); | |
91 | ||
92 | // Trailing bits | |
93 | if (last) | |
94 | FB_WRITEL( comp( FB_READL(src), FB_READL(dst), last), dst); | |
95 | } | |
96 | } else { | |
97 | unsigned long d0, d1; | |
98 | int m; | |
99 | // Different alignment for source and dest | |
100 | ||
101 | right = shift & (bits - 1); | |
102 | left = -shift & (bits - 1); | |
103 | ||
104 | if (dst_idx+n <= bits) { | |
105 | // Single destination word | |
106 | if (last) | |
107 | first &= last; | |
108 | if (shift > 0) { | |
109 | // Single source word | |
110 | FB_WRITEL( comp( FB_READL(src) >> right, FB_READL(dst), first), dst); | |
111 | } else if (src_idx+n <= bits) { | |
112 | // Single source word | |
113 | FB_WRITEL( comp(FB_READL(src) << left, FB_READL(dst), first), dst); | |
114 | } else { | |
115 | // 2 source words | |
116 | d0 = FB_READL(src++); | |
117 | d1 = FB_READL(src); | |
118 | FB_WRITEL( comp(d0<<left | d1>>right, FB_READL(dst), first), dst); | |
119 | } | |
120 | } else { | |
121 | // Multiple destination words | |
122 | /** We must always remember the last value read, because in case | |
123 | SRC and DST overlap bitwise (e.g. when moving just one pixel in | |
124 | 1bpp), we always collect one full long for DST and that might | |
125 | overlap with the current long from SRC. We store this value in | |
126 | 'd0'. */ | |
127 | d0 = FB_READL(src++); | |
128 | // Leading bits | |
129 | if (shift > 0) { | |
130 | // Single source word | |
131 | FB_WRITEL( comp(d0 >> right, FB_READL(dst), first), dst); | |
132 | dst++; | |
133 | n -= bits - dst_idx; | |
134 | } else { | |
135 | // 2 source words | |
136 | d1 = FB_READL(src++); | |
137 | FB_WRITEL( comp(d0<<left | d1>>right, FB_READL(dst), first), dst); | |
138 | d0 = d1; | |
139 | dst++; | |
140 | n -= bits - dst_idx; | |
141 | } | |
142 | ||
143 | // Main chunk | |
144 | m = n % bits; | |
145 | n /= bits; | |
146 | while (n >= 4) { | |
147 | d1 = FB_READL(src++); | |
148 | FB_WRITEL(d0 << left | d1 >> right, dst++); | |
149 | d0 = d1; | |
150 | d1 = FB_READL(src++); | |
151 | FB_WRITEL(d0 << left | d1 >> right, dst++); | |
152 | d0 = d1; | |
153 | d1 = FB_READL(src++); | |
154 | FB_WRITEL(d0 << left | d1 >> right, dst++); | |
155 | d0 = d1; | |
156 | d1 = FB_READL(src++); | |
157 | FB_WRITEL(d0 << left | d1 >> right, dst++); | |
158 | d0 = d1; | |
159 | n -= 4; | |
160 | } | |
161 | while (n--) { | |
162 | d1 = FB_READL(src++); | |
163 | FB_WRITEL(d0 << left | d1 >> right, dst++); | |
164 | d0 = d1; | |
165 | } | |
166 | ||
167 | // Trailing bits | |
168 | if (last) { | |
169 | if (m <= right) { | |
170 | // Single source word | |
171 | FB_WRITEL( comp(d0 << left, FB_READL(dst), last), dst); | |
172 | } else { | |
173 | // 2 source words | |
174 | d1 = FB_READL(src); | |
175 | FB_WRITEL( comp(d0<<left | d1>>right, FB_READL(dst), last), dst); | |
176 | } | |
177 | } | |
178 | } | |
179 | } | |
180 | } | |
181 | ||
182 | /* | |
183 | * Generic bitwise copy algorithm, operating backward | |
184 | */ | |
185 | ||
186 | static void | |
187 | bitcpy_rev(unsigned long __iomem *dst, int dst_idx, const unsigned long __iomem *src, | |
188 | int src_idx, int bits, unsigned n) | |
189 | { | |
190 | unsigned long first, last; | |
191 | int shift; | |
192 | ||
193 | dst += (n-1)/bits; | |
194 | src += (n-1)/bits; | |
195 | if ((n-1) % bits) { | |
196 | dst_idx += (n-1) % bits; | |
197 | dst += dst_idx >> (ffs(bits) - 1); | |
198 | dst_idx &= bits - 1; | |
199 | src_idx += (n-1) % bits; | |
200 | src += src_idx >> (ffs(bits) - 1); | |
201 | src_idx &= bits - 1; | |
202 | } | |
203 | ||
204 | shift = dst_idx-src_idx; | |
205 | ||
be0d9b6c AD |
206 | first = FB_SHIFT_LOW(~0UL, bits - 1 - dst_idx); |
207 | last = ~(FB_SHIFT_LOW(~0UL, bits - 1 - ((dst_idx-n) % bits))); | |
1da177e4 LT |
208 | |
209 | if (!shift) { | |
210 | // Same alignment for source and dest | |
211 | ||
212 | if ((unsigned long)dst_idx+1 >= n) { | |
213 | // Single word | |
214 | if (last) | |
215 | first &= last; | |
216 | FB_WRITEL( comp( FB_READL(src), FB_READL(dst), first), dst); | |
217 | } else { | |
218 | // Multiple destination words | |
219 | ||
220 | // Leading bits | |
221 | if (first != ~0UL) { | |
222 | FB_WRITEL( comp( FB_READL(src), FB_READL(dst), first), dst); | |
223 | dst--; | |
224 | src--; | |
225 | n -= dst_idx+1; | |
226 | } | |
227 | ||
228 | // Main chunk | |
229 | n /= bits; | |
230 | while (n >= 8) { | |
231 | FB_WRITEL(FB_READL(src--), dst--); | |
232 | FB_WRITEL(FB_READL(src--), dst--); | |
233 | FB_WRITEL(FB_READL(src--), dst--); | |
234 | FB_WRITEL(FB_READL(src--), dst--); | |
235 | FB_WRITEL(FB_READL(src--), dst--); | |
236 | FB_WRITEL(FB_READL(src--), dst--); | |
237 | FB_WRITEL(FB_READL(src--), dst--); | |
238 | FB_WRITEL(FB_READL(src--), dst--); | |
239 | n -= 8; | |
240 | } | |
241 | while (n--) | |
242 | FB_WRITEL(FB_READL(src--), dst--); | |
243 | ||
244 | // Trailing bits | |
245 | if (last) | |
246 | FB_WRITEL( comp( FB_READL(src), FB_READL(dst), last), dst); | |
247 | } | |
248 | } else { | |
249 | // Different alignment for source and dest | |
250 | ||
251 | int const left = -shift & (bits-1); | |
252 | int const right = shift & (bits-1); | |
253 | ||
254 | if ((unsigned long)dst_idx+1 >= n) { | |
255 | // Single destination word | |
256 | if (last) | |
257 | first &= last; | |
258 | if (shift < 0) { | |
259 | // Single source word | |
260 | FB_WRITEL( comp( FB_READL(src)<<left, FB_READL(dst), first), dst); | |
261 | } else if (1+(unsigned long)src_idx >= n) { | |
262 | // Single source word | |
263 | FB_WRITEL( comp( FB_READL(src)>>right, FB_READL(dst), first), dst); | |
264 | } else { | |
265 | // 2 source words | |
266 | FB_WRITEL( comp( (FB_READL(src)>>right | FB_READL(src-1)<<left), FB_READL(dst), first), dst); | |
267 | } | |
268 | } else { | |
269 | // Multiple destination words | |
270 | /** We must always remember the last value read, because in case | |
271 | SRC and DST overlap bitwise (e.g. when moving just one pixel in | |
272 | 1bpp), we always collect one full long for DST and that might | |
273 | overlap with the current long from SRC. We store this value in | |
274 | 'd0'. */ | |
275 | unsigned long d0, d1; | |
276 | int m; | |
277 | ||
278 | d0 = FB_READL(src--); | |
279 | // Leading bits | |
280 | if (shift < 0) { | |
281 | // Single source word | |
282 | FB_WRITEL( comp( (d0 << left), FB_READL(dst), first), dst); | |
283 | } else { | |
284 | // 2 source words | |
285 | d1 = FB_READL(src--); | |
286 | FB_WRITEL( comp( (d0>>right | d1<<left), FB_READL(dst), first), dst); | |
287 | d0 = d1; | |
288 | } | |
289 | dst--; | |
290 | n -= dst_idx+1; | |
291 | ||
292 | // Main chunk | |
293 | m = n % bits; | |
294 | n /= bits; | |
295 | while (n >= 4) { | |
296 | d1 = FB_READL(src--); | |
297 | FB_WRITEL(d0 >> right | d1 << left, dst--); | |
298 | d0 = d1; | |
299 | d1 = FB_READL(src--); | |
300 | FB_WRITEL(d0 >> right | d1 << left, dst--); | |
301 | d0 = d1; | |
302 | d1 = FB_READL(src--); | |
303 | FB_WRITEL(d0 >> right | d1 << left, dst--); | |
304 | d0 = d1; | |
305 | d1 = FB_READL(src--); | |
306 | FB_WRITEL(d0 >> right | d1 << left, dst--); | |
307 | d0 = d1; | |
308 | n -= 4; | |
309 | } | |
310 | while (n--) { | |
311 | d1 = FB_READL(src--); | |
312 | FB_WRITEL(d0 >> right | d1 << left, dst--); | |
313 | d0 = d1; | |
314 | } | |
315 | ||
316 | // Trailing bits | |
317 | if (last) { | |
318 | if (m <= left) { | |
319 | // Single source word | |
320 | FB_WRITEL( comp(d0 >> right, FB_READL(dst), last), dst); | |
321 | } else { | |
322 | // 2 source words | |
323 | d1 = FB_READL(src); | |
324 | FB_WRITEL( comp(d0>>right | d1<<left, FB_READL(dst), last), dst); | |
325 | } | |
326 | } | |
327 | } | |
328 | } | |
329 | } | |
330 | ||
331 | void cfb_copyarea(struct fb_info *p, const struct fb_copyarea *area) | |
332 | { | |
333 | u32 dx = area->dx, dy = area->dy, sx = area->sx, sy = area->sy; | |
334 | u32 height = area->height, width = area->width; | |
335 | unsigned long const bits_per_line = p->fix.line_length*8u; | |
336 | unsigned long __iomem *dst = NULL, *src = NULL; | |
337 | int bits = BITS_PER_LONG, bytes = bits >> 3; | |
338 | int dst_idx = 0, src_idx = 0, rev_copy = 0; | |
1da177e4 LT |
339 | |
340 | if (p->state != FBINFO_STATE_RUNNING) | |
341 | return; | |
342 | ||
1da177e4 LT |
343 | /* if the beginning of the target area might overlap with the end of |
344 | the source area, be have to copy the area reverse. */ | |
345 | if ((dy == sy && dx > sx) || (dy > sy)) { | |
346 | dy += height; | |
347 | sy += height; | |
348 | rev_copy = 1; | |
349 | } | |
350 | ||
351 | // split the base of the framebuffer into a long-aligned address and the | |
352 | // index of the first bit | |
353 | dst = src = (unsigned long __iomem *)((unsigned long)p->screen_base & ~(bytes-1)); | |
354 | dst_idx = src_idx = 8*((unsigned long)p->screen_base & (bytes-1)); | |
355 | // add offset of source and target area | |
356 | dst_idx += dy*bits_per_line + dx*p->var.bits_per_pixel; | |
357 | src_idx += sy*bits_per_line + sx*p->var.bits_per_pixel; | |
358 | ||
359 | if (p->fbops->fb_sync) | |
360 | p->fbops->fb_sync(p); | |
361 | ||
362 | if (rev_copy) { | |
363 | while (height--) { | |
364 | dst_idx -= bits_per_line; | |
365 | src_idx -= bits_per_line; | |
366 | dst += dst_idx >> (ffs(bits) - 1); | |
367 | dst_idx &= (bytes - 1); | |
368 | src += src_idx >> (ffs(bits) - 1); | |
369 | src_idx &= (bytes - 1); | |
370 | bitcpy_rev(dst, dst_idx, src, src_idx, bits, | |
371 | width*p->var.bits_per_pixel); | |
372 | } | |
373 | } else { | |
374 | while (height--) { | |
375 | dst += dst_idx >> (ffs(bits) - 1); | |
376 | dst_idx &= (bytes - 1); | |
377 | src += src_idx >> (ffs(bits) - 1); | |
378 | src_idx &= (bytes - 1); | |
379 | bitcpy(dst, dst_idx, src, src_idx, bits, | |
380 | width*p->var.bits_per_pixel); | |
381 | dst_idx += bits_per_line; | |
382 | src_idx += bits_per_line; | |
383 | } | |
384 | } | |
385 | } | |
386 | ||
387 | EXPORT_SYMBOL(cfb_copyarea); | |
388 | ||
389 | MODULE_AUTHOR("James Simmons <jsimmons@users.sf.net>"); | |
390 | MODULE_DESCRIPTION("Generic software accelerated copyarea"); | |
391 | MODULE_LICENSE("GPL"); | |
392 |