/draw/arch_port.c

https://github.com/Lafriks/mupdf · C · 486 lines · 446 code · 20 blank · 20 comment · 77 complexity · a33d1fce1b656cdc986274ebd815ee34 MD5 · raw file 

    

  1. #include "fitz.h"
    2. 

  2. 

  3. typedef unsigned char byte;
    4. 

  4. 

  5. /* These C implementations use SWAR (SIMD-within-a-register) techniques. */
    6. 

  6. 

  7. #if 0 /* TODO: move into porterduff.c functions */
    8. 

  8. 

  9. #define MASK 0xFF00FF00;
    10. 

  10. 

  11. static void
    12. 

  12. path_w4i1o4_32bit(byte *rgba,
    13. 

  13. 	byte * restrict src, byte cov, int len, byte * restrict dst)
    14. 

  14. {
    15. 

  15. 	/* COLOR * coverage + DST * (256-coverage) = (COLOR - DST)*coverage + DST*256 */
    16. 

  16. 	unsigned int *dst32 = (unsigned int *)(void *)dst;
    17. 

  17. 	int alpha = rgba[3];
    18. 

  18. 	unsigned int rb = rgba[0] | (rgba[2] << 16);
    19. 

  19. 	unsigned int ga = rgba[1] | 0xFF0000;
    20. 

  20. 

  21. 	if (alpha == 0)
    22. 

  22. 		return;
    23. 

  23. 

  24. 	if (alpha != 255)
    25. 

  25. 	{
    26. 

  26. 		alpha += alpha>>7; /* alpha is now in the 0...256 range */
    27. 

  27. 		while (len--)
    28. 

  28. 		{
    29. 

  29. 			unsigned int ca, drb, dga, crb, cga;
    30. 

  30. 			cov += *src; *src++ = 0;
    31. 

  31. 			ca = cov + (cov>>7); /* ca is in 0...256 range */
    32. 

  32. 			ca = (ca*alpha)>>8; /* ca is is in 0...256 range */
    33. 

  33. 			drb = *dst32++;
    34. 

  34. 			if (ca != 0)
    35. 

  35. 			{
    36. 

  36. 				dga = drb & MASK;
    37. 

  37. 				drb = (drb<<8) & MASK;
    38. 

  38. 				cga = ga - (dga>>8);
    39. 

  39. 				crb = rb - (drb>>8);
    40. 

  40. 				dga += cga * ca;
    41. 

  41. 				drb += crb * ca;
    42. 

  42. 				dga &= MASK;
    43. 

  43. 				drb &= MASK;
    44. 

  44. 				drb = dga | (drb>>8);
    45. 

  45. 				dst32[-1] = drb;
    46. 

  46. 			}
    47. 

  47. 		}
    48. 

  48. 	}
    49. 

  49. 	else
    50. 

  50. 	{
    51. 

  51. 		while (len--)
    52. 

  52. 		{
    53. 

  53. 			unsigned int ca, drb, dga, crb, cga;
    54. 

  54. 			cov += *src; *src++ = 0;
    55. 

  55. 			ca = cov + (cov>>7); /* ca is in 0...256 range */
    56. 

  56. 			drb = *dst32++;
    57. 

  57. 			if (ca == 0)
    58. 

  58. 				continue;
    59. 

  59. 			if (ca == 255)
    60. 

  60. 			{
    61. 

  61. 				drb = (ga<<8) | rb;
    62. 

  62. 			}
    63. 

  63. 			else
    64. 

  64. 			{
    65. 

  65. 				dga = drb & MASK;
    66. 

  66. 				drb = (drb<<8) & MASK;
    67. 

  67. 				cga = ga - (dga>>8);
    68. 

  68. 				crb = rb - (drb>>8);
    69. 

  69. 				dga += cga * ca;
    70. 

  70. 				drb += crb * ca;
    71. 

  71. 				dga &= MASK;
    72. 

  72. 				drb &= MASK;
    73. 

  73. 				drb = dga |(drb>>8);
    74. 

  74. 			}
    75. 

  75. 			dst32[-1] = drb;
    76. 

  76. 		}
    77. 

  77. 	}
    78. 

  78. }
    79. 

  79. 

  80. static void
    81. 

  81. text_w4i1o4_32bit(byte *rgba,
    82. 

  82. 	byte * restrict src, int srcw,
    83. 

  83. 	byte * restrict dst, int dstw, int w0, int h)
    84. 

  84. {
    85. 

  85. 	unsigned int *dst32 = (unsigned int *)(void *)dst;
    86. 

  86. 	unsigned int alpha = rgba[3];
    87. 

  87. 	unsigned int rb = rgba[0] | (rgba[2] << 16);
    88. 

  88. 	unsigned int ga = rgba[1] | 0xFF0000;
    89. 

  89. 

  90. 	if (alpha == 0)
    91. 

  91. 		return;
    92. 

  92. 

  93. 	srcw -= w0;
    94. 

  94. 	dstw = (dstw>>2)-w0;
    95. 

  95. 

  96. 	if (alpha != 255)
    97. 

  97. 	{
    98. 

  98. 		alpha += alpha>>7; /* alpha is now in the 0...256 range */
    99. 

  99. 		while (h--)
    100. 

  100. 		{
    101. 

  101. 			int w = w0;
    102. 

  102. 			while (w--)
    103. 

  103. 			{
    104. 

  104. 				unsigned int ca, drb, dga, crb, cga;
    105. 

  105. 				ca = *src++;
    106. 

  106. 				drb = *dst32++;
    107. 

  107. 				ca += ca>>7;
    108. 

  108. 				ca = (ca*alpha)>>8;
    109. 

  109. 				if (ca == 0)
    110. 

  110. 					continue;
    111. 

  111. 				dga = drb & MASK;
    112. 

  112. 				drb = (drb<<8) & MASK;
    113. 

  113. 				cga = ga - (dga>>8);
    114. 

  114. 				crb = rb - (drb>>8);
    115. 

  115. 				dga += cga * ca;
    116. 

  116. 				drb += crb * ca;
    117. 

  117. 				dga &= MASK;
    118. 

  118. 				drb &= MASK;
    119. 

  119. 				drb = dga | (drb>>8);
    120. 

  120. 				dst32[-1] = drb;
    121. 

  121. 			}
    122. 

  122. 			src += srcw;
    123. 

  123. 			dst32 += dstw;
    124. 

  124. 		}
    125. 

  125. 	}
    126. 

  126. 	else
    127. 

  127. 	{
    128. 

  128. 		while (h--)
    129. 

  129. 		{
    130. 

  130. 			int w = w0;
    131. 

  131. 			while (w--)
    132. 

  132. 			{
    133. 

  133. 				unsigned int ca, drb, dga, crb, cga;
    134. 

  134. 				ca = *src++;
    135. 

  135. 				drb = *dst32++;
    136. 

  136. 				ca += ca>>7;
    137. 

  137. 				if (ca == 0)
    138. 

  138. 					continue;
    139. 

  139. 				dga = drb & MASK;
    140. 

  140. 				drb = (drb<<8) & MASK;
    141. 

  141. 				cga = ga - (dga>>8);
    142. 

  142. 				crb = rb - (drb>>8);
    143. 

  143. 				dga += cga * ca;
    144. 

  144. 				drb += crb * ca;
    145. 

  145. 				dga &= MASK;
    146. 

  146. 				drb &= MASK;
    147. 

  147. 				drb = dga | (drb>>8);
    148. 

  148. 				dst32[-1] = drb;
    149. 

  149. 			}
    150. 

  150. 			src += srcw;
    151. 

  151. 			dst32 += dstw;
    152. 

  152. 		}
    153. 

  153. 	}
    154. 

  154. }
    155. 

  155. 

  156. static void
    157. 

  157. img_4o4_32bit(byte * restrict src, byte cov, int len, byte * restrict dst,
    158. 

  158. 	fz_pixmap *image, int u, int v, int fa, int fb)
    159. 

  159. {
    160. 

  160. 	unsigned int *dst32 = (unsigned int *)(void *)dst;
    161. 

  161. 	unsigned int *samples = (unsigned int *)(void *)image->samples;
    162. 

  162. 	int w = image->w;
    163. 

  163. 	int h = image->h-1;
    164. 

  164. 

  165. 	while (len--)
    166. 

  166. 	{
    167. 

  167. 		unsigned int a, a1, d, d1;
    168. 

  168. 		int sa;
    169. 

  169. 		cov += *src; *src = 0; src++;
    170. 

  170. 		/* (a,a1) = sampleargb(samples, w, h, u, v, argb); */
    171. 

  171. 		{
    172. 

  172. 			int ui, ui1, vi, vi1, ud, vd;
    173. 

  173. 			unsigned int b, b1, c, c1;
    174. 

  174. 			ui1 = 1;
    175. 

  175. 			ui = u >> 16;
    176. 

  176. 			if (ui < 0)
    177. 

  177. 			{
    178. 

  178. 				ui = 0;
    179. 

  179. 				ui1 = 0;
    180. 

  180. 			}
    181. 

  181. 			else if (ui >= w-1)
    182. 

  182. 			{
    183. 

  183. 				ui = w-1;
    184. 

  184. 				ui1 = 0;
    185. 

  185. 			}
    186. 

  186. 			vi1 = w;
    187. 

  187. 			vi = v >> 16;
    188. 

  188. 			if (vi < 0)
    189. 

  189. 			{
    190. 

  190. 				vi = 0;
    191. 

  191. 				vi1 = 0;
    192. 

  192. 			}
    193. 

  193. 			else if (vi >= h)
    194. 

  194. 			{
    195. 

  195. 				vi = h;
    196. 

  196. 				vi1 = 0;
    197. 

  197. 			}
    198. 

  198. 			ui += vi*w;
    199. 

  199. 			a = samples[ui];
    200. 

  200. 			b = samples[ui + ui1];
    201. 

  201. 			c = samples[ui + vi1];
    202. 

  202. 			d = samples[ui + ui1 + vi1];
    203. 

  203. 			ud = (u>>8) & 0xFF;
    204. 

  204. 			vd = (v>>8) & 0xFF;
    205. 

  205. 			ud = FZ_EXPAND(ud);
    206. 

  206. 			vd = FZ_EXPAND(vd);
    207. 

  207. 			/* (a,a1) = blend(a,b,ud) */
    208. 

  208. 			a1 = a & MASK;
    209. 

  209. 			a = (a<<8) & MASK;
    210. 

  210. 			b1 = (b>>8) & ~MASK;
    211. 

  211. 			b = b & ~MASK;
    212. 

  212. 			a = ((b -(a >>8)) * ud + a ) & MASK;
    213. 

  213. 			a1 = ((b1-(a1>>8)) * ud + a1) & MASK;
    214. 

  214. 			/* (c,c1) = blend(c,d,ud) */
    215. 

  215. 			c1 = c & MASK;
    216. 

  216. 			c = (c<<8) & MASK;
    217. 

  217. 			d1 = (d>>8) & ~MASK;
    218. 

  218. 			d = d & ~MASK;
    219. 

  219. 			c = ((d -(c >>8)) * ud + c ) & MASK;
    220. 

  220. 			c1 = ((d1-(c1>>8)) * ud + c1) & MASK;
    221. 

  221. 			/* (a,a1) = blend((a,a1),(c,c1),vd) */
    222. 

  222. 			a = (((c >>8)-(a >>8)) * vd + a ) & MASK;
    223. 

  223. 			a1 = (((c1>>8)-(a1>>8)) * vd + a1) & MASK;
    224. 

  224. 		}
    225. 

  225. 		sa = (a1>>24);
    226. 

  226. 		sa = FZ_COMBINE(FZ_EXPAND(sa), FZ_EXPAND(cov));
    227. 

  227. 		a1 |= 0xFF000000;
    228. 

  228. 		d = *dst32++;
    229. 

  229. 		d1 = d & MASK;
    230. 

  230. 		d = (d<<8) & MASK;
    231. 

  231. 		a = (((a >>8)-(d >>8)) * sa + d ) & MASK;
    232. 

  232. 		a1 = (((a1>>8)-(d1>>8)) * sa + d1) & MASK;
    233. 

  233. 		dst32[-1] = (a>>8) | a1;
    234. 

  234. 		u += fa;
    235. 

  235. 		v += fb;
    236. 

  236. 	}
    237. 

  237. }
    238. 

  238. 

  239. static void
    240. 

  240. img_w4i1o4_32bit(byte *rgba, byte * restrict src, byte cov, int len,
    241. 

  241. 	byte * restrict dst, fz_pixmap *image, int u, int v, int fa, int fb)
    242. 

  242. {
    243. 

  243. 	byte *samples = image->samples;
    244. 

  244. 	int w = image->w;
    245. 

  245. 	int h = image->h-1;
    246. 

  246. 	int alpha = FZ_EXPAND(rgba[3]);
    247. 

  247. 	unsigned int rb = rgba[0] | (rgba[2] << 16);
    248. 

  248. 	unsigned int ga = rgba[1] | 0xFF0000;
    249. 

  249. 	unsigned int *dst32 = (unsigned int *)(void *)dst;
    250. 

  250. 

  251. 	if (alpha == 0)
    252. 

  252. 		return;
    253. 

  253. 	if (alpha != 256)
    254. 

  254. 	{
    255. 

  255. 		while (len--)
    256. 

  256. 		{
    257. 

  257. 			unsigned int ca, drb, dga, crb, cga;
    258. 

  258. 			unsigned int a, b;
    259. 

  259. 			cov += *src; *src = 0; src++;
    260. 

  260. 			drb = *dst32++;
    261. 

  261. 			ca = FZ_COMBINE(FZ_EXPAND(cov), alpha);
    262. 

  262. 			if (ca != 0)
    263. 

  263. 			{
    264. 

  264. 				int ui, ui1, vi, vi1, ud, vd;
    265. 

  265. 				/* a = samplemask(samples, w, h, u, v); */
    266. 

  266. 				ui1 = 1;
    267. 

  267. 				ui = u >> 16;
    268. 

  268. 				if (ui < 0)
    269. 

  269. 				{
    270. 

  270. 					ui = 0;
    271. 

  271. 					ui1 = 0;
    272. 

  272. 				}
    273. 

  273. 				else if (ui >= w-1)
    274. 

  274. 				{
    275. 

  275. 					ui = w-1;
    276. 

  276. 					ui1 = 0;
    277. 

  277. 				}
    278. 

  278. 				vi1 = w;
    279. 

  279. 				vi = v >> 16;
    280. 

  280. 				if (vi < 0)
    281. 

  281. 				{
    282. 

  282. 					vi = 0;
    283. 

  283. 					vi1 = 0;
    284. 

  284. 				}
    285. 

  285. 				else if (vi >= h)
    286. 

  286. 				{
    287. 

  287. 					vi = h;
    288. 

  288. 					vi1 = 0;
    289. 

  289. 				}
    290. 

  290. 				ui += vi*w;
    291. 

  291. 				a = samples[ui];
    292. 

  292. 				b = samples[ui + ui1];
    293. 

  293. 				a |= samples[ui + vi1]<<16;
    294. 

  294. 				b |= samples[ui + ui1 + vi1]<<16;
    295. 

  295. 				ud = (u>>8) & 0xFF;
    296. 

  296. 				vd = (v>>8) & 0xFF;
    297. 

  297. 				ud = FZ_EXPAND(ud);
    298. 

  298. 				vd = FZ_EXPAND(vd);
    299. 

  299. 				/* a = blend(a,b,ud) */
    300. 

  300. 				a = ((b-a) * ud + (a<<8)) & MASK;
    301. 

  301. 				/* a = blend(a,a>>16,vd) */
    302. 

  302. 				a = (((a>>24)-(a>>8)) * vd + a);
    303. 

  303. 				a = (a>>8) & 0xFF;
    304. 

  304. 				ca = FZ_COMBINE(ca, FZ_EXPAND(a));
    305. 

  305. 			}
    306. 

  306. 			if (ca != 0)
    307. 

  307. 			{
    308. 

  308. 				dga = drb & MASK;
    309. 

  309. 				drb = (drb<<8) & MASK;
    310. 

  310. 				cga = ga - (dga>>8);
    311. 

  311. 				crb = rb - (drb>>8);
    312. 

  312. 				dga += cga * ca;
    313. 

  313. 				drb += crb * ca;
    314. 

  314. 				dga &= MASK;
    315. 

  315. 				drb &= MASK;
    316. 

  316. 				drb = dga | (drb>>8);
    317. 

  317. 				dst32[-1] = drb;
    318. 

  318. 			}
    319. 

  319. 			u += fa;
    320. 

  320. 			v += fb;
    321. 

  321. 		}
    322. 

  322. 	}
    323. 

  323. 	else
    324. 

  324. 	{
    325. 

  325. 		while (len--)
    326. 

  326. 		{
    327. 

  327. 			unsigned int ca, drb, dga, crb, cga;
    328. 

  328. 			unsigned int a, b;
    329. 

  329. 			cov += *src; *src = 0; src++;
    330. 

  330. 			drb = *dst32++;
    331. 

  331. 			if (cov != 0)
    332. 

  332. 			{
    333. 

  333. 				int ui, ui1, vi, vi1, ud, vd;
    334. 

  334. 				/* a = samplemask(samples, w, h, u, v); */
    335. 

  335. 				ui1 = 1;
    336. 

  336. 				ui = u >> 16;
    337. 

  337. 				if (ui < 0)
    338. 

  338. 				{
    339. 

  339. 					ui = 0;
    340. 

  340. 					ui1 = 0;
    341. 

  341. 				}
    342. 

  342. 				else if (ui >= w-1)
    343. 

  343. 				{
    344. 

  344. 					ui = w-1;
    345. 

  345. 					ui1 = 0;
    346. 

  346. 				}
    347. 

  347. 				vi1 = w;
    348. 

  348. 				vi = v >> 16;
    349. 

  349. 				if (vi < 0)
    350. 

  350. 				{
    351. 

  351. 					vi = 0;
    352. 

  352. 					vi1 = 0;
    353. 

  353. 				}
    354. 

  354. 				else if (vi >= h)
    355. 

  355. 				{
    356. 

  356. 					vi = h;
    357. 

  357. 					vi1 = 0;
    358. 

  358. 				}
    359. 

  359. 				ui += vi*w;
    360. 

  360. 				a = samples[ui];
    361. 

  361. 				b = samples[ui + ui1];
    362. 

  362. 				a |= samples[ui + vi1]<<16;
    363. 

  363. 				b |= samples[ui + ui1 + vi1]<<16;
    364. 

  364. 				ud = (u>>8) & 0xFF;
    365. 

  365. 				vd = (v>>8) & 0xFF;
    366. 

  366. 				ud = FZ_EXPAND(ud);
    367. 

  367. 				vd = FZ_EXPAND(vd);
    368. 

  368. 				/* a = blend(a,b,ud) */
    369. 

  369. 				a = ((b-a) * ud + (a<<8)) & MASK;
    370. 

  370. 				/* a = blend(a,a>>16,vd) */
    371. 

  371. 				a = (((a>>24)-(a>>8)) * vd + a);
    372. 

  372. 				a = (a>>8) & 0xFF;
    373. 

  373. 				ca = FZ_COMBINE(FZ_EXPAND(cov),FZ_EXPAND(a));
    374. 

  374. 				if (ca != 0)
    375. 

  375. 				{
    376. 

  376. 					if (ca == 256)
    377. 

  377. 					{
    378. 

  378. 						drb = (ga<<8) | rb;
    379. 

  379. 					}
    380. 

  380. 					else
    381. 

  381. 					{
    382. 

  382. 						dga = drb & MASK;
    383. 

  383. 						drb = (drb<<8) & MASK;
    384. 

  384. 						cga = ga - (dga>>8);
    385. 

  385. 						crb = rb - (drb>>8);
    386. 

  386. 						dga += cga * ca;
    387. 

  387. 						drb += crb * ca;
    388. 

  388. 						dga &= MASK;
    389. 

  389. 						drb &= MASK;
    390. 

  390. 						drb = dga | (drb>>8);
    391. 

  391. 					}
    392. 

  392. 					dst32[-1] = drb;
    393. 

  393. 				}
    394. 

  394. 			}
    395. 

  395. 			u += fa;
    396. 

  396. 			v += fb;
    397. 

  397. 		}
    398. 

  398. 	}
    399. 

  399. }
    400. 

  400. 

  401. static void
    402. 

  402. img_1o1_32bit(byte * restrict src, byte cov, int len, byte * restrict dst,
    403. 

  403. 	fz_pixmap *image, int u, int v, int fa, int fb)
    404. 

  404. {
    405. 

  405. 	byte *samples = image->samples;
    406. 

  406. 	int w = image->w;
    407. 

  407. 	int h = image->h-1;
    408. 

  408. 

  409. 	while (len--)
    410. 

  410. 	{
    411. 

  411. 		unsigned int a, b;
    412. 

  412. 		cov += *src; *src = 0; src++;
    413. 

  413. 		if (cov != 0)
    414. 

  414. 		{
    415. 

  415. 			int ui, ui1, vi, vi1, ud, vd;
    416. 

  416. 			/* sa = samplemask(samples, w, h, u, v); */
    417. 

  417. 			ui1 = 1;
    418. 

  418. 			ui = u >> 16;
    419. 

  419. 			if (ui < 0)
    420. 

  420. 			{
    421. 

  421. 				ui = 0;
    422. 

  422. 				ui1 = 0;
    423. 

  423. 			}
    424. 

  424. 			else if (ui >= w-1)
    425. 

  425. 			{
    426. 

  426. 				ui = w-1;
    427. 

  427. 				ui1 = 0;
    428. 

  428. 			}
    429. 

  429. 			vi1 = w;
    430. 

  430. 			vi = v >> 16;
    431. 

  431. 			if (vi < 0)
    432. 

  432. 			{
    433. 

  433. 				vi = 0;
    434. 

  434. 				vi1 = 0;
    435. 

  435. 			}
    436. 

  436. 			else if (vi >= h)
    437. 

  437. 			{
    438. 

  438. 				vi = h;
    439. 

  439. 				vi1 = 0;
    440. 

  440. 			}
    441. 

  441. 			ui += vi*w;
    442. 

  442. 			a = samples[ui];
    443. 

  443. 			b = samples[ui + ui1];
    444. 

  444. 			a |= samples[ui + vi1]<<16;
    445. 

  445. 			b |= samples[ui + ui1 + vi1]<<16;
    446. 

  446. 			ud = (u>>8) & 0xFF;
    447. 

  447. 			vd = (v>>8) & 0xFF;
    448. 

  448. 			ud = FZ_EXPAND(ud);
    449. 

  449. 			vd = FZ_EXPAND(vd);
    450. 

  450. 			/* a = blend(a,b,ud) */
    451. 

  451. 			a = ((b-a) * ud + (a<<8)) & MASK;
    452. 

  452. 			/* a = blend(a,a>>16,vd) */
    453. 

  453. 			a = (((a>>24)-(a>>8)) * vd + a);
    454. 

  454. 			a = (a>>8) & 0xFF;
    455. 

  455. 			a = FZ_COMBINE(FZ_EXPAND(a), FZ_EXPAND(cov));
    456. 

  456. 			if (a != 0)
    457. 

  457. 			{
    458. 

  458. 				if (a == 256)
    459. 

  459. 					dst[0] = 255;
    460. 

  460. 				else
    461. 

  461. 					dst[0] = FZ_BLEND(255, dst[0], a);
    462. 

  462. 			}
    463. 

  463. 		}
    464. 

  464. 		dst++;
    465. 

  465. 		u += fa;
    466. 

  466. 		v += fb;
    467. 

  467. 	}
    468. 

  468. }
    469. 

  469. 

  470. #endif
    471. 

  471. 

  472. void fz_accelerate(void)
    473. 

  473. {
    474. 

  474. 	if (sizeof(int) == 4 && sizeof(unsigned int) == 4 && !fz_is_big_endian())
    475. 

  475. 	{
    476. 

  476. //		fz_path_w4i1o4 = path_w4i1o4_32bit;
    477. 

  477. //		fz_text_w4i1o4 = text_w4i1o4_32bit;
    478. 

  478. //		fz_img_4o4 = img_4o4_32bit;
    479. 

  479. //		fz_img_w4i1o4 = img_w4i1o4_32bit;
    480. 

  480. //		fz_img_1o1 = img_1o1_32bit;
    481. 

  481. 	}
    482. 

  482. 

  483. #ifdef HAVE_CPUDEP
    484. 

  484. 	fz_accelerate_arch();
    485. 

  485. #endif
    486. 

  486. }
    487.