e14743d1 |
1 | /* |
2 | SDL - Simple DirectMedia Layer |
3 | Copyright (C) 1997-2009 Sam Lantinga |
4 | |
5 | This library is free software; you can redistribute it and/or |
6 | modify it under the terms of the GNU Lesser General Public |
7 | License as published by the Free Software Foundation; either |
8 | version 2.1 of the License, or (at your option) any later version. |
9 | |
10 | This library is distributed in the hope that it will be useful, |
11 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
12 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
13 | Lesser General Public License for more details. |
14 | |
15 | You should have received a copy of the GNU Lesser General Public |
16 | License along with this library; if not, write to the Free Software |
17 | Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA |
18 | |
19 | Sam Lantinga |
20 | slouken@libsdl.org |
21 | */ |
22 | #include "SDL_config.h" |
23 | |
24 | #include "SDL_video.h" |
25 | #include "SDL_blit.h" |
26 | |
27 | /* |
28 | In Visual C, VC6 has mmintrin.h in the "Processor Pack" add-on. |
29 | Checking if _mm_free is #defined in malloc.h is is the only way to |
30 | determine if the Processor Pack is installed, as far as I can tell. |
31 | */ |
32 | |
33 | #if SDL_ASSEMBLY_ROUTINES |
34 | # if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__)) |
35 | # define MMX_ASMBLIT 1 |
36 | # define GCC_ASMBLIT 1 |
37 | # elif defined(_MSC_VER) && defined(_M_IX86) |
38 | # if (_MSC_VER <= 1200) |
39 | # include <malloc.h> |
40 | # if defined(_mm_free) |
41 | # define HAVE_MMINTRIN_H 1 |
42 | # endif |
43 | # else /* Visual Studio > VC6 always has mmintrin.h */ |
44 | # define HAVE_MMINTRIN_H 1 |
45 | # endif |
46 | # if HAVE_MMINTRIN_H |
47 | # define MMX_ASMBLIT 1 |
48 | # define MSVC_ASMBLIT 1 |
49 | # endif |
50 | # endif |
51 | #endif /* SDL_ASSEMBLY_ROUTINES */ |
52 | |
53 | /* Function to check the CPU flags */ |
54 | #include "SDL_cpuinfo.h" |
55 | #if GCC_ASMBLIT |
56 | #include "mmx.h" |
57 | #elif MSVC_ASMBLIT |
58 | #include <mmintrin.h> |
59 | #include <mm3dnow.h> |
60 | #endif |
61 | |
62 | /* Functions to perform alpha blended blitting */ |
63 | |
a1f34081 |
64 | #ifdef __ARM_NEON__ |
65 | |
66 | /* NEON optimized blitter callers */ |
67 | #define make_neon_caller(name, neon_name) \ |
68 | extern void neon_name(void *dst, const void *src, int count); \ |
69 | static void name(SDL_BlitInfo *info) \ |
70 | { \ |
71 | int width = info->d_width; \ |
72 | int height = info->d_height; \ |
73 | Uint8 *src = info->s_pixels; \ |
74 | Uint8 *dst = info->d_pixels; \ |
75 | int srcskip = info->s_skip; \ |
76 | int dstskip = info->d_skip; \ |
77 | \ |
78 | while ( height-- ) { \ |
79 | neon_name(dst, src, width); \ |
80 | src += width * 4 + srcskip; \ |
81 | dst += width * 4 + dstskip; \ |
82 | } \ |
83 | } |
84 | |
bdfa6989 |
85 | #define make_neon_callerS(name, neon_name) \ |
86 | extern void neon_name(void *dst, const void *src, int count, unsigned int alpha); \ |
87 | static void name(SDL_BlitInfo *info) \ |
88 | { \ |
89 | int width = info->d_width; \ |
90 | int height = info->d_height; \ |
91 | Uint8 *src = info->s_pixels; \ |
92 | Uint8 *dst = info->d_pixels; \ |
93 | int srcskip = info->s_skip; \ |
94 | int dstskip = info->d_skip; \ |
95 | unsigned alpha = info->src->alpha;\ |
96 | \ |
97 | while ( height-- ) { \ |
98 | neon_name(dst, src, width, alpha); \ |
99 | src += width * 4 + srcskip; \ |
100 | dst += width * 4 + dstskip; \ |
101 | } \ |
102 | } |
103 | |
a1f34081 |
104 | make_neon_caller(BlitABGRtoXRGBalpha_neon, neon_ABGRtoXRGBalpha) |
105 | make_neon_caller(BlitARGBtoXRGBalpha_neon, neon_ARGBtoXRGBalpha) |
bdfa6989 |
106 | make_neon_callerS(BlitABGRtoXRGBalphaS_neon, neon_ABGRtoXRGBalphaS) |
107 | make_neon_callerS(BlitARGBtoXRGBalphaS_neon, neon_ARGBtoXRGBalphaS) |
a1f34081 |
108 | |
109 | #endif /* __ARM_NEON__ */ |
110 | |
e14743d1 |
111 | /* N->1 blending with per-surface alpha */ |
112 | static void BlitNto1SurfaceAlpha(SDL_BlitInfo *info) |
113 | { |
114 | int width = info->d_width; |
115 | int height = info->d_height; |
116 | Uint8 *src = info->s_pixels; |
117 | int srcskip = info->s_skip; |
118 | Uint8 *dst = info->d_pixels; |
119 | int dstskip = info->d_skip; |
120 | Uint8 *palmap = info->table; |
121 | SDL_PixelFormat *srcfmt = info->src; |
122 | SDL_PixelFormat *dstfmt = info->dst; |
123 | int srcbpp = srcfmt->BytesPerPixel; |
124 | |
125 | const unsigned A = srcfmt->alpha; |
126 | |
127 | while ( height-- ) { |
128 | DUFFS_LOOP4( |
129 | { |
130 | Uint32 Pixel; |
131 | unsigned sR; |
132 | unsigned sG; |
133 | unsigned sB; |
134 | unsigned dR; |
135 | unsigned dG; |
136 | unsigned dB; |
137 | DISEMBLE_RGB(src, srcbpp, srcfmt, Pixel, sR, sG, sB); |
138 | dR = dstfmt->palette->colors[*dst].r; |
139 | dG = dstfmt->palette->colors[*dst].g; |
140 | dB = dstfmt->palette->colors[*dst].b; |
141 | ALPHA_BLEND(sR, sG, sB, A, dR, dG, dB); |
142 | dR &= 0xff; |
143 | dG &= 0xff; |
144 | dB &= 0xff; |
145 | /* Pack RGB into 8bit pixel */ |
146 | if ( palmap == NULL ) { |
147 | *dst =((dR>>5)<<(3+2))| |
148 | ((dG>>5)<<(2))| |
149 | ((dB>>6)<<(0)); |
150 | } else { |
151 | *dst = palmap[((dR>>5)<<(3+2))| |
152 | ((dG>>5)<<(2)) | |
153 | ((dB>>6)<<(0))]; |
154 | } |
155 | dst++; |
156 | src += srcbpp; |
157 | }, |
158 | width); |
159 | src += srcskip; |
160 | dst += dstskip; |
161 | } |
162 | } |
163 | |
164 | /* N->1 blending with pixel alpha */ |
165 | static void BlitNto1PixelAlpha(SDL_BlitInfo *info) |
166 | { |
167 | int width = info->d_width; |
168 | int height = info->d_height; |
169 | Uint8 *src = info->s_pixels; |
170 | int srcskip = info->s_skip; |
171 | Uint8 *dst = info->d_pixels; |
172 | int dstskip = info->d_skip; |
173 | Uint8 *palmap = info->table; |
174 | SDL_PixelFormat *srcfmt = info->src; |
175 | SDL_PixelFormat *dstfmt = info->dst; |
176 | int srcbpp = srcfmt->BytesPerPixel; |
177 | |
178 | /* FIXME: fix alpha bit field expansion here too? */ |
179 | while ( height-- ) { |
180 | DUFFS_LOOP4( |
181 | { |
182 | Uint32 Pixel; |
183 | unsigned sR; |
184 | unsigned sG; |
185 | unsigned sB; |
186 | unsigned sA; |
187 | unsigned dR; |
188 | unsigned dG; |
189 | unsigned dB; |
190 | DISEMBLE_RGBA(src,srcbpp,srcfmt,Pixel,sR,sG,sB,sA); |
191 | dR = dstfmt->palette->colors[*dst].r; |
192 | dG = dstfmt->palette->colors[*dst].g; |
193 | dB = dstfmt->palette->colors[*dst].b; |
194 | ALPHA_BLEND(sR, sG, sB, sA, dR, dG, dB); |
195 | dR &= 0xff; |
196 | dG &= 0xff; |
197 | dB &= 0xff; |
198 | /* Pack RGB into 8bit pixel */ |
199 | if ( palmap == NULL ) { |
200 | *dst =((dR>>5)<<(3+2))| |
201 | ((dG>>5)<<(2))| |
202 | ((dB>>6)<<(0)); |
203 | } else { |
204 | *dst = palmap[((dR>>5)<<(3+2))| |
205 | ((dG>>5)<<(2)) | |
206 | ((dB>>6)<<(0)) ]; |
207 | } |
208 | dst++; |
209 | src += srcbpp; |
210 | }, |
211 | width); |
212 | src += srcskip; |
213 | dst += dstskip; |
214 | } |
215 | } |
216 | |
217 | /* colorkeyed N->1 blending with per-surface alpha */ |
218 | static void BlitNto1SurfaceAlphaKey(SDL_BlitInfo *info) |
219 | { |
220 | int width = info->d_width; |
221 | int height = info->d_height; |
222 | Uint8 *src = info->s_pixels; |
223 | int srcskip = info->s_skip; |
224 | Uint8 *dst = info->d_pixels; |
225 | int dstskip = info->d_skip; |
226 | Uint8 *palmap = info->table; |
227 | SDL_PixelFormat *srcfmt = info->src; |
228 | SDL_PixelFormat *dstfmt = info->dst; |
229 | int srcbpp = srcfmt->BytesPerPixel; |
230 | Uint32 ckey = srcfmt->colorkey; |
231 | |
232 | const int A = srcfmt->alpha; |
233 | |
234 | while ( height-- ) { |
235 | DUFFS_LOOP( |
236 | { |
237 | Uint32 Pixel; |
238 | unsigned sR; |
239 | unsigned sG; |
240 | unsigned sB; |
241 | unsigned dR; |
242 | unsigned dG; |
243 | unsigned dB; |
244 | DISEMBLE_RGB(src, srcbpp, srcfmt, Pixel, sR, sG, sB); |
245 | if ( Pixel != ckey ) { |
246 | dR = dstfmt->palette->colors[*dst].r; |
247 | dG = dstfmt->palette->colors[*dst].g; |
248 | dB = dstfmt->palette->colors[*dst].b; |
249 | ALPHA_BLEND(sR, sG, sB, A, dR, dG, dB); |
250 | dR &= 0xff; |
251 | dG &= 0xff; |
252 | dB &= 0xff; |
253 | /* Pack RGB into 8bit pixel */ |
254 | if ( palmap == NULL ) { |
255 | *dst =((dR>>5)<<(3+2))| |
256 | ((dG>>5)<<(2)) | |
257 | ((dB>>6)<<(0)); |
258 | } else { |
259 | *dst = palmap[((dR>>5)<<(3+2))| |
260 | ((dG>>5)<<(2)) | |
261 | ((dB>>6)<<(0)) ]; |
262 | } |
263 | } |
264 | dst++; |
265 | src += srcbpp; |
266 | }, |
267 | width); |
268 | src += srcskip; |
269 | dst += dstskip; |
270 | } |
271 | } |
272 | |
273 | #if GCC_ASMBLIT |
274 | /* fast RGB888->(A)RGB888 blending with surface alpha=128 special case */ |
275 | static void BlitRGBtoRGBSurfaceAlpha128MMX(SDL_BlitInfo *info) |
276 | { |
277 | int width = info->d_width; |
278 | int height = info->d_height; |
279 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
280 | int srcskip = info->s_skip >> 2; |
281 | Uint32 *dstp = (Uint32 *)info->d_pixels; |
282 | int dstskip = info->d_skip >> 2; |
283 | Uint32 dalpha = info->dst->Amask; |
284 | Uint64 load; |
285 | |
286 | load = 0x00fefefe00fefefeULL;/* alpha128 mask */ |
287 | movq_m2r(load, mm4); /* alpha128 mask -> mm4 */ |
288 | load = 0x0001010100010101ULL;/* !alpha128 mask */ |
289 | movq_m2r(load, mm3); /* !alpha128 mask -> mm3 */ |
290 | movd_m2r(dalpha, mm7); /* dst alpha mask */ |
291 | punpckldq_r2r(mm7, mm7); /* dst alpha mask | dst alpha mask -> mm7 */ |
292 | while(height--) { |
293 | DUFFS_LOOP_DOUBLE2( |
294 | { |
295 | Uint32 s = *srcp++; |
296 | Uint32 d = *dstp; |
297 | *dstp++ = ((((s & 0x00fefefe) + (d & 0x00fefefe)) >> 1) |
298 | + (s & d & 0x00010101)) | dalpha; |
299 | },{ |
300 | movq_m2r((*dstp), mm2);/* 2 x dst -> mm2(ARGBARGB) */ |
301 | movq_r2r(mm2, mm6); /* 2 x dst -> mm6(ARGBARGB) */ |
302 | |
303 | movq_m2r((*srcp), mm1);/* 2 x src -> mm1(ARGBARGB) */ |
304 | movq_r2r(mm1, mm5); /* 2 x src -> mm5(ARGBARGB) */ |
305 | |
306 | pand_r2r(mm4, mm6); /* dst & mask -> mm6 */ |
307 | pand_r2r(mm4, mm5); /* src & mask -> mm5 */ |
308 | paddd_r2r(mm6, mm5); /* mm6 + mm5 -> mm5 */ |
309 | pand_r2r(mm1, mm2); /* src & dst -> mm2 */ |
310 | psrld_i2r(1, mm5); /* mm5 >> 1 -> mm5 */ |
311 | pand_r2r(mm3, mm2); /* mm2 & !mask -> mm2 */ |
312 | paddd_r2r(mm5, mm2); /* mm5 + mm2 -> mm2 */ |
313 | |
314 | por_r2r(mm7, mm2); /* mm7(full alpha) | mm2 -> mm2 */ |
315 | movq_r2m(mm2, (*dstp));/* mm2 -> 2 x dst pixels */ |
316 | dstp += 2; |
317 | srcp += 2; |
318 | }, width); |
319 | srcp += srcskip; |
320 | dstp += dstskip; |
321 | } |
322 | emms(); |
323 | } |
324 | |
325 | /* fast RGB888->(A)RGB888 blending with surface alpha */ |
326 | static void BlitRGBtoRGBSurfaceAlphaMMX(SDL_BlitInfo *info) |
327 | { |
328 | SDL_PixelFormat* df = info->dst; |
329 | unsigned alpha = info->src->alpha; |
330 | |
331 | if (alpha == 128 && (df->Rmask | df->Gmask | df->Bmask) == 0x00FFFFFF) { |
332 | /* only call a128 version when R,G,B occupy lower bits */ |
333 | BlitRGBtoRGBSurfaceAlpha128MMX(info); |
334 | } else { |
335 | int width = info->d_width; |
336 | int height = info->d_height; |
337 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
338 | int srcskip = info->s_skip >> 2; |
339 | Uint32 *dstp = (Uint32 *)info->d_pixels; |
340 | int dstskip = info->d_skip >> 2; |
341 | |
342 | pxor_r2r(mm5, mm5); /* 0 -> mm5 */ |
343 | /* form the alpha mult */ |
344 | movd_m2r(alpha, mm4); /* 0000000A -> mm4 */ |
345 | punpcklwd_r2r(mm4, mm4); /* 00000A0A -> mm4 */ |
346 | punpckldq_r2r(mm4, mm4); /* 0A0A0A0A -> mm4 */ |
347 | alpha = (0xff << df->Rshift) | (0xff << df->Gshift) | (0xff << df->Bshift); |
348 | movd_m2r(alpha, mm0); /* 00000FFF -> mm0 */ |
349 | punpcklbw_r2r(mm0, mm0); /* 00FFFFFF -> mm0 */ |
350 | pand_r2r(mm0, mm4); /* 0A0A0A0A -> mm4, minus 1 chan */ |
351 | /* at this point mm4 can be 000A0A0A or 0A0A0A00 or another combo */ |
352 | movd_m2r(df->Amask, mm7); /* dst alpha mask */ |
353 | punpckldq_r2r(mm7, mm7); /* dst alpha mask | dst alpha mask -> mm7 */ |
354 | |
355 | while(height--) { |
356 | DUFFS_LOOP_DOUBLE2({ |
357 | /* One Pixel Blend */ |
358 | movd_m2r((*srcp), mm1);/* src(ARGB) -> mm1 (0000ARGB)*/ |
359 | movd_m2r((*dstp), mm2);/* dst(ARGB) -> mm2 (0000ARGB)*/ |
360 | punpcklbw_r2r(mm5, mm1); /* 0A0R0G0B -> mm1(src) */ |
361 | punpcklbw_r2r(mm5, mm2); /* 0A0R0G0B -> mm2(dst) */ |
362 | |
363 | psubw_r2r(mm2, mm1);/* src - dst -> mm1 */ |
364 | pmullw_r2r(mm4, mm1); /* mm1 * alpha -> mm1 */ |
365 | psrlw_i2r(8, mm1); /* mm1 >> 8 -> mm1 */ |
366 | paddb_r2r(mm1, mm2); /* mm1 + mm2(dst) -> mm2 */ |
367 | |
368 | packuswb_r2r(mm5, mm2); /* ARGBARGB -> mm2 */ |
369 | por_r2r(mm7, mm2); /* mm7(full alpha) | mm2 -> mm2 */ |
370 | movd_r2m(mm2, *dstp);/* mm2 -> pixel */ |
371 | ++srcp; |
372 | ++dstp; |
373 | },{ |
374 | /* Two Pixels Blend */ |
375 | movq_m2r((*srcp), mm0);/* 2 x src -> mm0(ARGBARGB)*/ |
376 | movq_m2r((*dstp), mm2);/* 2 x dst -> mm2(ARGBARGB) */ |
377 | movq_r2r(mm0, mm1); /* 2 x src -> mm1(ARGBARGB) */ |
378 | movq_r2r(mm2, mm6); /* 2 x dst -> mm6(ARGBARGB) */ |
379 | |
380 | punpcklbw_r2r(mm5, mm0); /* low - 0A0R0G0B -> mm0(src1) */ |
381 | punpckhbw_r2r(mm5, mm1); /* high - 0A0R0G0B -> mm1(src2) */ |
382 | punpcklbw_r2r(mm5, mm2); /* low - 0A0R0G0B -> mm2(dst1) */ |
383 | punpckhbw_r2r(mm5, mm6); /* high - 0A0R0G0B -> mm6(dst2) */ |
384 | |
385 | psubw_r2r(mm2, mm0);/* src1 - dst1 -> mm0 */ |
386 | pmullw_r2r(mm4, mm0); /* mm0 * alpha -> mm0 */ |
387 | psrlw_i2r(8, mm0); /* mm0 >> 8 -> mm1 */ |
388 | paddb_r2r(mm0, mm2); /* mm0 + mm2(dst1) -> mm2 */ |
389 | |
390 | psubw_r2r(mm6, mm1);/* src2 - dst2 -> mm1 */ |
391 | pmullw_r2r(mm4, mm1); /* mm1 * alpha -> mm1 */ |
392 | psrlw_i2r(8, mm1); /* mm1 >> 8 -> mm1 */ |
393 | paddb_r2r(mm1, mm6); /* mm1 + mm6(dst2) -> mm6 */ |
394 | |
395 | packuswb_r2r(mm6, mm2); /* ARGBARGB -> mm2 */ |
396 | por_r2r(mm7, mm2); /* mm7(dst alpha) | mm2 -> mm2 */ |
397 | |
398 | movq_r2m(mm2, *dstp);/* mm2 -> 2 x pixel */ |
399 | |
400 | srcp += 2; |
401 | dstp += 2; |
402 | }, width); |
403 | srcp += srcskip; |
404 | dstp += dstskip; |
405 | } |
406 | emms(); |
407 | } |
408 | } |
409 | |
410 | /* fast ARGB888->(A)RGB888 blending with pixel alpha */ |
411 | static void BlitRGBtoRGBPixelAlphaMMX(SDL_BlitInfo *info) |
412 | { |
413 | int width = info->d_width; |
414 | int height = info->d_height; |
415 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
416 | int srcskip = info->s_skip >> 2; |
417 | Uint32 *dstp = (Uint32 *)info->d_pixels; |
418 | int dstskip = info->d_skip >> 2; |
419 | SDL_PixelFormat* sf = info->src; |
420 | Uint32 amask = sf->Amask; |
421 | |
422 | pxor_r2r(mm6, mm6); /* 0 -> mm6 */ |
423 | /* form multiplication mask */ |
424 | movd_m2r(sf->Amask, mm7); /* 0000F000 -> mm7 */ |
425 | punpcklbw_r2r(mm7, mm7); /* FF000000 -> mm7 */ |
426 | pcmpeqb_r2r(mm0, mm0); /* FFFFFFFF -> mm0 */ |
427 | movq_r2r(mm0, mm3); /* FFFFFFFF -> mm3 (for later) */ |
428 | pxor_r2r(mm0, mm7); /* 00FFFFFF -> mm7 (mult mask) */ |
429 | /* form channel masks */ |
430 | movq_r2r(mm7, mm0); /* 00FFFFFF -> mm0 */ |
431 | packsswb_r2r(mm6, mm0); /* 00000FFF -> mm0 (channel mask) */ |
432 | packsswb_r2r(mm6, mm3); /* 0000FFFF -> mm3 */ |
433 | pxor_r2r(mm0, mm3); /* 0000F000 -> mm3 (~channel mask) */ |
434 | /* get alpha channel shift */ |
435 | __asm__ __volatile__ ( |
436 | "movd %0, %%mm5" |
437 | : : "rm" ((Uint32) sf->Ashift) ); /* Ashift -> mm5 */ |
438 | |
439 | while(height--) { |
440 | DUFFS_LOOP4({ |
441 | Uint32 alpha = *srcp & amask; |
442 | /* FIXME: Here we special-case opaque alpha since the |
443 | compositioning used (>>8 instead of /255) doesn't handle |
444 | it correctly. Also special-case alpha=0 for speed? |
445 | Benchmark this! */ |
446 | if(alpha == 0) { |
447 | /* do nothing */ |
448 | } else if(alpha == amask) { |
449 | /* opaque alpha -- copy RGB, keep dst alpha */ |
450 | /* using MMX here to free up regular registers for other things */ |
451 | movd_m2r((*srcp), mm1);/* src(ARGB) -> mm1 (0000ARGB)*/ |
452 | movd_m2r((*dstp), mm2);/* dst(ARGB) -> mm2 (0000ARGB)*/ |
453 | pand_r2r(mm0, mm1); /* src & chanmask -> mm1 */ |
454 | pand_r2r(mm3, mm2); /* dst & ~chanmask -> mm2 */ |
455 | por_r2r(mm1, mm2); /* src | dst -> mm2 */ |
456 | movd_r2m(mm2, (*dstp)); /* mm2 -> dst */ |
457 | } else { |
458 | movd_m2r((*srcp), mm1);/* src(ARGB) -> mm1 (0000ARGB)*/ |
459 | punpcklbw_r2r(mm6, mm1); /* 0A0R0G0B -> mm1 */ |
460 | |
461 | movd_m2r((*dstp), mm2);/* dst(ARGB) -> mm2 (0000ARGB)*/ |
462 | punpcklbw_r2r(mm6, mm2); /* 0A0R0G0B -> mm2 */ |
463 | |
464 | __asm__ __volatile__ ( |
465 | "movd %0, %%mm4" |
466 | : : "r" (alpha) ); /* 0000A000 -> mm4 */ |
467 | psrld_r2r(mm5, mm4); /* mm4 >> mm5 -> mm4 (0000000A) */ |
468 | punpcklwd_r2r(mm4, mm4); /* 00000A0A -> mm4 */ |
469 | punpcklwd_r2r(mm4, mm4); /* 0A0A0A0A -> mm4 */ |
470 | pand_r2r(mm7, mm4); /* 000A0A0A -> mm4, preserve dst alpha on add */ |
471 | |
472 | /* blend */ |
473 | psubw_r2r(mm2, mm1);/* src - dst -> mm1 */ |
474 | pmullw_r2r(mm4, mm1); /* mm1 * alpha -> mm1 */ |
475 | psrlw_i2r(8, mm1); /* mm1 >> 8 -> mm1(000R0G0B) */ |
476 | paddb_r2r(mm1, mm2); /* mm1 + mm2(dst) -> mm2 */ |
477 | |
478 | packuswb_r2r(mm6, mm2); /* 0000ARGB -> mm2 */ |
479 | movd_r2m(mm2, *dstp);/* mm2 -> dst */ |
480 | } |
481 | ++srcp; |
482 | ++dstp; |
483 | }, width); |
484 | srcp += srcskip; |
485 | dstp += dstskip; |
486 | } |
487 | emms(); |
488 | } |
489 | /* End GCC_ASMBLIT */ |
490 | |
491 | #elif MSVC_ASMBLIT |
492 | /* fast RGB888->(A)RGB888 blending with surface alpha=128 special case */ |
493 | static void BlitRGBtoRGBSurfaceAlpha128MMX(SDL_BlitInfo *info) |
494 | { |
495 | int width = info->d_width; |
496 | int height = info->d_height; |
497 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
498 | int srcskip = info->s_skip >> 2; |
499 | Uint32 *dstp = (Uint32 *)info->d_pixels; |
500 | int dstskip = info->d_skip >> 2; |
501 | Uint32 dalpha = info->dst->Amask; |
502 | |
503 | __m64 src1, src2, dst1, dst2, lmask, hmask, dsta; |
504 | |
505 | hmask = _mm_set_pi32(0x00fefefe, 0x00fefefe); /* alpha128 mask -> hmask */ |
506 | lmask = _mm_set_pi32(0x00010101, 0x00010101); /* !alpha128 mask -> lmask */ |
507 | dsta = _mm_set_pi32(dalpha, dalpha); /* dst alpha mask -> dsta */ |
508 | |
509 | while (height--) { |
510 | int n = width; |
511 | if ( n & 1 ) { |
512 | Uint32 s = *srcp++; |
513 | Uint32 d = *dstp; |
514 | *dstp++ = ((((s & 0x00fefefe) + (d & 0x00fefefe)) >> 1) |
515 | + (s & d & 0x00010101)) | dalpha; |
516 | n--; |
517 | } |
518 | |
519 | for (n >>= 1; n > 0; --n) { |
520 | dst1 = *(__m64*)dstp; /* 2 x dst -> dst1(ARGBARGB) */ |
521 | dst2 = dst1; /* 2 x dst -> dst2(ARGBARGB) */ |
522 | |
523 | src1 = *(__m64*)srcp; /* 2 x src -> src1(ARGBARGB) */ |
524 | src2 = src1; /* 2 x src -> src2(ARGBARGB) */ |
525 | |
526 | dst2 = _mm_and_si64(dst2, hmask); /* dst & mask -> dst2 */ |
527 | src2 = _mm_and_si64(src2, hmask); /* src & mask -> src2 */ |
528 | src2 = _mm_add_pi32(src2, dst2); /* dst2 + src2 -> src2 */ |
529 | src2 = _mm_srli_pi32(src2, 1); /* src2 >> 1 -> src2 */ |
530 | |
531 | dst1 = _mm_and_si64(dst1, src1); /* src & dst -> dst1 */ |
532 | dst1 = _mm_and_si64(dst1, lmask); /* dst1 & !mask -> dst1 */ |
533 | dst1 = _mm_add_pi32(dst1, src2); /* src2 + dst1 -> dst1 */ |
534 | dst1 = _mm_or_si64(dst1, dsta); /* dsta(full alpha) | dst1 -> dst1 */ |
535 | |
536 | *(__m64*)dstp = dst1; /* dst1 -> 2 x dst pixels */ |
537 | dstp += 2; |
538 | srcp += 2; |
539 | } |
540 | |
541 | srcp += srcskip; |
542 | dstp += dstskip; |
543 | } |
544 | _mm_empty(); |
545 | } |
546 | |
547 | /* fast RGB888->(A)RGB888 blending with surface alpha */ |
548 | static void BlitRGBtoRGBSurfaceAlphaMMX(SDL_BlitInfo *info) |
549 | { |
550 | SDL_PixelFormat* df = info->dst; |
551 | Uint32 chanmask = df->Rmask | df->Gmask | df->Bmask; |
552 | unsigned alpha = info->src->alpha; |
553 | |
554 | if (alpha == 128 && (df->Rmask | df->Gmask | df->Bmask) == 0x00FFFFFF) { |
555 | /* only call a128 version when R,G,B occupy lower bits */ |
556 | BlitRGBtoRGBSurfaceAlpha128MMX(info); |
557 | } else { |
558 | int width = info->d_width; |
559 | int height = info->d_height; |
560 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
561 | int srcskip = info->s_skip >> 2; |
562 | Uint32 *dstp = (Uint32 *)info->d_pixels; |
563 | int dstskip = info->d_skip >> 2; |
564 | Uint32 dalpha = df->Amask; |
565 | Uint32 amult; |
566 | |
567 | __m64 src1, src2, dst1, dst2, mm_alpha, mm_zero, dsta; |
568 | |
569 | mm_zero = _mm_setzero_si64(); /* 0 -> mm_zero */ |
570 | /* form the alpha mult */ |
571 | amult = alpha | (alpha << 8); |
572 | amult = amult | (amult << 16); |
573 | chanmask = (0xff << df->Rshift) | (0xff << df->Gshift) | (0xff << df->Bshift); |
574 | mm_alpha = _mm_set_pi32(0, amult & chanmask); /* 0000AAAA -> mm_alpha, minus 1 chan */ |
575 | mm_alpha = _mm_unpacklo_pi8(mm_alpha, mm_zero); /* 0A0A0A0A -> mm_alpha, minus 1 chan */ |
576 | /* at this point mm_alpha can be 000A0A0A or 0A0A0A00 or another combo */ |
577 | dsta = _mm_set_pi32(dalpha, dalpha); /* dst alpha mask -> dsta */ |
578 | |
579 | while (height--) { |
580 | int n = width; |
581 | if (n & 1) { |
582 | /* One Pixel Blend */ |
583 | src2 = _mm_cvtsi32_si64(*srcp); /* src(ARGB) -> src2 (0000ARGB)*/ |
584 | src2 = _mm_unpacklo_pi8(src2, mm_zero); /* 0A0R0G0B -> src2 */ |
585 | |
586 | dst1 = _mm_cvtsi32_si64(*dstp); /* dst(ARGB) -> dst1 (0000ARGB)*/ |
587 | dst1 = _mm_unpacklo_pi8(dst1, mm_zero); /* 0A0R0G0B -> dst1 */ |
588 | |
589 | src2 = _mm_sub_pi16(src2, dst1); /* src2 - dst2 -> src2 */ |
590 | src2 = _mm_mullo_pi16(src2, mm_alpha); /* src2 * alpha -> src2 */ |
591 | src2 = _mm_srli_pi16(src2, 8); /* src2 >> 8 -> src2 */ |
592 | dst1 = _mm_add_pi8(src2, dst1); /* src2 + dst1 -> dst1 */ |
593 | |
594 | dst1 = _mm_packs_pu16(dst1, mm_zero); /* 0000ARGB -> dst1 */ |
595 | dst1 = _mm_or_si64(dst1, dsta); /* dsta | dst1 -> dst1 */ |
596 | *dstp = _mm_cvtsi64_si32(dst1); /* dst1 -> pixel */ |
597 | |
598 | ++srcp; |
599 | ++dstp; |
600 | |
601 | n--; |
602 | } |
603 | |
604 | for (n >>= 1; n > 0; --n) { |
605 | /* Two Pixels Blend */ |
606 | src1 = *(__m64*)srcp; /* 2 x src -> src1(ARGBARGB)*/ |
607 | src2 = src1; /* 2 x src -> src2(ARGBARGB) */ |
608 | src1 = _mm_unpacklo_pi8(src1, mm_zero); /* low - 0A0R0G0B -> src1 */ |
609 | src2 = _mm_unpackhi_pi8(src2, mm_zero); /* high - 0A0R0G0B -> src2 */ |
610 | |
611 | dst1 = *(__m64*)dstp;/* 2 x dst -> dst1(ARGBARGB) */ |
612 | dst2 = dst1; /* 2 x dst -> dst2(ARGBARGB) */ |
613 | dst1 = _mm_unpacklo_pi8(dst1, mm_zero); /* low - 0A0R0G0B -> dst1 */ |
614 | dst2 = _mm_unpackhi_pi8(dst2, mm_zero); /* high - 0A0R0G0B -> dst2 */ |
615 | |
616 | src1 = _mm_sub_pi16(src1, dst1);/* src1 - dst1 -> src1 */ |
617 | src1 = _mm_mullo_pi16(src1, mm_alpha); /* src1 * alpha -> src1 */ |
618 | src1 = _mm_srli_pi16(src1, 8); /* src1 >> 8 -> src1 */ |
619 | dst1 = _mm_add_pi8(src1, dst1); /* src1 + dst1(dst1) -> dst1 */ |
620 | |
621 | src2 = _mm_sub_pi16(src2, dst2);/* src2 - dst2 -> src2 */ |
622 | src2 = _mm_mullo_pi16(src2, mm_alpha); /* src2 * alpha -> src2 */ |
623 | src2 = _mm_srli_pi16(src2, 8); /* src2 >> 8 -> src2 */ |
624 | dst2 = _mm_add_pi8(src2, dst2); /* src2 + dst2(dst2) -> dst2 */ |
625 | |
626 | dst1 = _mm_packs_pu16(dst1, dst2); /* 0A0R0G0B(res1), 0A0R0G0B(res2) -> dst1(ARGBARGB) */ |
627 | dst1 = _mm_or_si64(dst1, dsta); /* dsta | dst1 -> dst1 */ |
628 | |
629 | *(__m64*)dstp = dst1; /* dst1 -> 2 x pixel */ |
630 | |
631 | srcp += 2; |
632 | dstp += 2; |
633 | } |
634 | srcp += srcskip; |
635 | dstp += dstskip; |
636 | } |
637 | _mm_empty(); |
638 | } |
639 | } |
640 | |
641 | /* fast ARGB888->(A)RGB888 blending with pixel alpha */ |
642 | static void BlitRGBtoRGBPixelAlphaMMX(SDL_BlitInfo *info) |
643 | { |
644 | int width = info->d_width; |
645 | int height = info->d_height; |
646 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
647 | int srcskip = info->s_skip >> 2; |
648 | Uint32 *dstp = (Uint32 *)info->d_pixels; |
649 | int dstskip = info->d_skip >> 2; |
650 | SDL_PixelFormat* sf = info->src; |
651 | Uint32 chanmask = sf->Rmask | sf->Gmask | sf->Bmask; |
652 | Uint32 amask = sf->Amask; |
653 | Uint32 ashift = sf->Ashift; |
654 | Uint64 multmask; |
655 | |
656 | __m64 src1, dst1, mm_alpha, mm_zero, dmask; |
657 | |
658 | mm_zero = _mm_setzero_si64(); /* 0 -> mm_zero */ |
659 | multmask = ~(0xFFFFi64 << (ashift * 2)); |
660 | dmask = *(__m64*) &multmask; /* dst alpha mask -> dmask */ |
661 | |
662 | while(height--) { |
663 | DUFFS_LOOP4({ |
664 | Uint32 alpha = *srcp & amask; |
665 | if (alpha == 0) { |
666 | /* do nothing */ |
667 | } else if (alpha == amask) { |
668 | /* opaque alpha -- copy RGB, keep dst alpha */ |
669 | *dstp = (*srcp & chanmask) | (*dstp & ~chanmask); |
670 | } else { |
671 | src1 = _mm_cvtsi32_si64(*srcp); /* src(ARGB) -> src1 (0000ARGB)*/ |
672 | src1 = _mm_unpacklo_pi8(src1, mm_zero); /* 0A0R0G0B -> src1 */ |
673 | |
674 | dst1 = _mm_cvtsi32_si64(*dstp); /* dst(ARGB) -> dst1 (0000ARGB)*/ |
675 | dst1 = _mm_unpacklo_pi8(dst1, mm_zero); /* 0A0R0G0B -> dst1 */ |
676 | |
677 | mm_alpha = _mm_cvtsi32_si64(alpha); /* alpha -> mm_alpha (0000000A) */ |
678 | mm_alpha = _mm_srli_si64(mm_alpha, ashift); /* mm_alpha >> ashift -> mm_alpha(0000000A) */ |
679 | mm_alpha = _mm_unpacklo_pi16(mm_alpha, mm_alpha); /* 00000A0A -> mm_alpha */ |
680 | mm_alpha = _mm_unpacklo_pi32(mm_alpha, mm_alpha); /* 0A0A0A0A -> mm_alpha */ |
681 | mm_alpha = _mm_and_si64(mm_alpha, dmask); /* 000A0A0A -> mm_alpha, preserve dst alpha on add */ |
682 | |
683 | /* blend */ |
684 | src1 = _mm_sub_pi16(src1, dst1);/* src1 - dst1 -> src1 */ |
685 | src1 = _mm_mullo_pi16(src1, mm_alpha); /* (src1 - dst1) * alpha -> src1 */ |
686 | src1 = _mm_srli_pi16(src1, 8); /* src1 >> 8 -> src1(000R0G0B) */ |
687 | dst1 = _mm_add_pi8(src1, dst1); /* src1 + dst1 -> dst1(0A0R0G0B) */ |
688 | dst1 = _mm_packs_pu16(dst1, mm_zero); /* 0000ARGB -> dst1 */ |
689 | |
690 | *dstp = _mm_cvtsi64_si32(dst1); /* dst1 -> pixel */ |
691 | } |
692 | ++srcp; |
693 | ++dstp; |
694 | }, width); |
695 | srcp += srcskip; |
696 | dstp += dstskip; |
697 | } |
698 | _mm_empty(); |
699 | } |
700 | /* End MSVC_ASMBLIT */ |
701 | |
702 | #endif /* GCC_ASMBLIT, MSVC_ASMBLIT */ |
703 | |
704 | #if SDL_ALTIVEC_BLITTERS |
705 | #if __MWERKS__ |
706 | #pragma altivec_model on |
707 | #endif |
708 | #if HAVE_ALTIVEC_H |
709 | #include <altivec.h> |
710 | #endif |
711 | #include <assert.h> |
712 | |
713 | #if (defined(__MACOSX__) && (__GNUC__ < 4)) |
714 | #define VECUINT8_LITERAL(a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p) \ |
715 | (vector unsigned char) ( a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p ) |
716 | #define VECUINT16_LITERAL(a,b,c,d,e,f,g,h) \ |
717 | (vector unsigned short) ( a,b,c,d,e,f,g,h ) |
718 | #else |
719 | #define VECUINT8_LITERAL(a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p) \ |
720 | (vector unsigned char) { a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p } |
721 | #define VECUINT16_LITERAL(a,b,c,d,e,f,g,h) \ |
722 | (vector unsigned short) { a,b,c,d,e,f,g,h } |
723 | #endif |
724 | |
725 | #define UNALIGNED_PTR(x) (((size_t) x) & 0x0000000F) |
726 | #define VECPRINT(msg, v) do { \ |
727 | vector unsigned int tmpvec = (vector unsigned int)(v); \ |
728 | unsigned int *vp = (unsigned int *)&tmpvec; \ |
729 | printf("%s = %08X %08X %08X %08X\n", msg, vp[0], vp[1], vp[2], vp[3]); \ |
730 | } while (0) |
731 | |
732 | /* the permuation vector that takes the high bytes out of all the appropriate shorts |
733 | (vector unsigned char)( |
734 | 0x00, 0x10, 0x02, 0x12, |
735 | 0x04, 0x14, 0x06, 0x16, |
736 | 0x08, 0x18, 0x0A, 0x1A, |
737 | 0x0C, 0x1C, 0x0E, 0x1E ); |
738 | */ |
739 | #define VEC_MERGE_PERMUTE() (vec_add(vec_lvsl(0, (int*)NULL), (vector unsigned char)vec_splat_u16(0x0F))) |
740 | #define VEC_U32_24() (vec_add(vec_splat_u32(12), vec_splat_u32(12))) |
741 | #define VEC_ALPHA_MASK() ((vector unsigned char)vec_sl((vector unsigned int)vec_splat_s8(-1), VEC_U32_24())) |
742 | #define VEC_ALIGNER(src) ((UNALIGNED_PTR(src)) \ |
743 | ? vec_lvsl(0, src) \ |
744 | : vec_add(vec_lvsl(8, src), vec_splat_u8(8))) |
745 | |
746 | |
747 | #define VEC_MULTIPLY_ALPHA(vs, vd, valpha, mergePermute, v1_16, v8_16) do { \ |
748 | /* vtemp1 contains source AAGGAAGGAAGGAAGG */ \ |
749 | vector unsigned short vtemp1 = vec_mule(vs, valpha); \ |
750 | /* vtemp2 contains source RRBBRRBBRRBBRRBB */ \ |
751 | vector unsigned short vtemp2 = vec_mulo(vs, valpha); \ |
752 | /* valpha2 is 255-alpha */ \ |
753 | vector unsigned char valpha2 = vec_nor(valpha, valpha); \ |
754 | /* vtemp3 contains dest AAGGAAGGAAGGAAGG */ \ |
755 | vector unsigned short vtemp3 = vec_mule(vd, valpha2); \ |
756 | /* vtemp4 contains dest RRBBRRBBRRBBRRBB */ \ |
757 | vector unsigned short vtemp4 = vec_mulo(vd, valpha2); \ |
758 | /* add source and dest */ \ |
759 | vtemp1 = vec_add(vtemp1, vtemp3); \ |
760 | vtemp2 = vec_add(vtemp2, vtemp4); \ |
761 | /* vtemp1 = (vtemp1 + 1) + ((vtemp1 + 1) >> 8) */ \ |
762 | vtemp1 = vec_add(vtemp1, v1_16); \ |
763 | vtemp3 = vec_sr(vtemp1, v8_16); \ |
764 | vtemp1 = vec_add(vtemp1, vtemp3); \ |
765 | /* vtemp2 = (vtemp2 + 1) + ((vtemp2 + 1) >> 8) */ \ |
766 | vtemp2 = vec_add(vtemp2, v1_16); \ |
767 | vtemp4 = vec_sr(vtemp2, v8_16); \ |
768 | vtemp2 = vec_add(vtemp2, vtemp4); \ |
769 | /* (>>8) and get ARGBARGBARGBARGB */ \ |
770 | vd = (vector unsigned char)vec_perm(vtemp1, vtemp2, mergePermute); \ |
771 | } while (0) |
772 | |
773 | /* Calculate the permute vector used for 32->32 swizzling */ |
774 | static vector unsigned char calc_swizzle32(const SDL_PixelFormat *srcfmt, |
775 | const SDL_PixelFormat *dstfmt) |
776 | { |
777 | /* |
778 | * We have to assume that the bits that aren't used by other |
779 | * colors is alpha, and it's one complete byte, since some formats |
780 | * leave alpha with a zero mask, but we should still swizzle the bits. |
781 | */ |
782 | /* ARGB */ |
783 | const static struct SDL_PixelFormat default_pixel_format = { |
784 | NULL, 0, 0, |
785 | 0, 0, 0, 0, |
786 | 16, 8, 0, 24, |
787 | 0x00FF0000, 0x0000FF00, 0x000000FF, 0xFF000000, |
788 | 0, 0}; |
789 | if (!srcfmt) { |
790 | srcfmt = &default_pixel_format; |
791 | } |
792 | if (!dstfmt) { |
793 | dstfmt = &default_pixel_format; |
794 | } |
795 | const vector unsigned char plus = VECUINT8_LITERAL |
796 | ( 0x00, 0x00, 0x00, 0x00, |
797 | 0x04, 0x04, 0x04, 0x04, |
798 | 0x08, 0x08, 0x08, 0x08, |
799 | 0x0C, 0x0C, 0x0C, 0x0C ); |
800 | vector unsigned char vswiz; |
801 | vector unsigned int srcvec; |
802 | #define RESHIFT(X) (3 - ((X) >> 3)) |
803 | Uint32 rmask = RESHIFT(srcfmt->Rshift) << (dstfmt->Rshift); |
804 | Uint32 gmask = RESHIFT(srcfmt->Gshift) << (dstfmt->Gshift); |
805 | Uint32 bmask = RESHIFT(srcfmt->Bshift) << (dstfmt->Bshift); |
806 | Uint32 amask; |
807 | /* Use zero for alpha if either surface doesn't have alpha */ |
808 | if (dstfmt->Amask) { |
809 | amask = ((srcfmt->Amask) ? RESHIFT(srcfmt->Ashift) : 0x10) << (dstfmt->Ashift); |
810 | } else { |
811 | amask = 0x10101010 & ((dstfmt->Rmask | dstfmt->Gmask | dstfmt->Bmask) ^ 0xFFFFFFFF); |
812 | } |
813 | #undef RESHIFT |
814 | ((unsigned int *)(char*)&srcvec)[0] = (rmask | gmask | bmask | amask); |
815 | vswiz = vec_add(plus, (vector unsigned char)vec_splat(srcvec, 0)); |
816 | return(vswiz); |
817 | } |
818 | |
819 | static void Blit32to565PixelAlphaAltivec(SDL_BlitInfo *info) |
820 | { |
821 | int height = info->d_height; |
822 | Uint8 *src = (Uint8 *)info->s_pixels; |
823 | int srcskip = info->s_skip; |
824 | Uint8 *dst = (Uint8 *)info->d_pixels; |
825 | int dstskip = info->d_skip; |
826 | SDL_PixelFormat *srcfmt = info->src; |
827 | |
828 | vector unsigned char v0 = vec_splat_u8(0); |
829 | vector unsigned short v8_16 = vec_splat_u16(8); |
830 | vector unsigned short v1_16 = vec_splat_u16(1); |
831 | vector unsigned short v2_16 = vec_splat_u16(2); |
832 | vector unsigned short v3_16 = vec_splat_u16(3); |
833 | vector unsigned int v8_32 = vec_splat_u32(8); |
834 | vector unsigned int v16_32 = vec_add(v8_32, v8_32); |
835 | vector unsigned short v3f = VECUINT16_LITERAL( |
836 | 0x003f, 0x003f, 0x003f, 0x003f, |
837 | 0x003f, 0x003f, 0x003f, 0x003f); |
838 | vector unsigned short vfc = VECUINT16_LITERAL( |
839 | 0x00fc, 0x00fc, 0x00fc, 0x00fc, |
840 | 0x00fc, 0x00fc, 0x00fc, 0x00fc); |
841 | |
842 | /* |
843 | 0x10 - 0x1f is the alpha |
844 | 0x00 - 0x0e evens are the red |
845 | 0x01 - 0x0f odds are zero |
846 | */ |
847 | vector unsigned char vredalpha1 = VECUINT8_LITERAL( |
848 | 0x10, 0x00, 0x01, 0x01, |
849 | 0x10, 0x02, 0x01, 0x01, |
850 | 0x10, 0x04, 0x01, 0x01, |
851 | 0x10, 0x06, 0x01, 0x01 |
852 | ); |
853 | vector unsigned char vredalpha2 = (vector unsigned char)( |
854 | vec_add((vector unsigned int)vredalpha1, vec_sl(v8_32, v16_32)) |
855 | ); |
856 | /* |
857 | 0x00 - 0x0f is ARxx ARxx ARxx ARxx |
858 | 0x11 - 0x0f odds are blue |
859 | */ |
860 | vector unsigned char vblue1 = VECUINT8_LITERAL( |
861 | 0x00, 0x01, 0x02, 0x11, |
862 | 0x04, 0x05, 0x06, 0x13, |
863 | 0x08, 0x09, 0x0a, 0x15, |
864 | 0x0c, 0x0d, 0x0e, 0x17 |
865 | ); |
866 | vector unsigned char vblue2 = (vector unsigned char)( |
867 | vec_add((vector unsigned int)vblue1, v8_32) |
868 | ); |
869 | /* |
870 | 0x00 - 0x0f is ARxB ARxB ARxB ARxB |
871 | 0x10 - 0x0e evens are green |
872 | */ |
873 | vector unsigned char vgreen1 = VECUINT8_LITERAL( |
874 | 0x00, 0x01, 0x10, 0x03, |
875 | 0x04, 0x05, 0x12, 0x07, |
876 | 0x08, 0x09, 0x14, 0x0b, |
877 | 0x0c, 0x0d, 0x16, 0x0f |
878 | ); |
879 | vector unsigned char vgreen2 = (vector unsigned char)( |
880 | vec_add((vector unsigned int)vgreen1, vec_sl(v8_32, v8_32)) |
881 | ); |
882 | vector unsigned char vgmerge = VECUINT8_LITERAL( |
883 | 0x00, 0x02, 0x00, 0x06, |
884 | 0x00, 0x0a, 0x00, 0x0e, |
885 | 0x00, 0x12, 0x00, 0x16, |
886 | 0x00, 0x1a, 0x00, 0x1e); |
887 | vector unsigned char mergePermute = VEC_MERGE_PERMUTE(); |
888 | vector unsigned char vpermute = calc_swizzle32(srcfmt, NULL); |
889 | vector unsigned char valphaPermute = vec_and(vec_lvsl(0, (int *)NULL), vec_splat_u8(0xC)); |
890 | |
891 | vector unsigned short vf800 = (vector unsigned short)vec_splat_u8(-7); |
892 | vf800 = vec_sl(vf800, vec_splat_u16(8)); |
893 | |
894 | while(height--) { |
895 | int extrawidth; |
896 | vector unsigned char valigner; |
897 | vector unsigned char vsrc; |
898 | vector unsigned char voverflow; |
899 | int width = info->d_width; |
900 | |
901 | #define ONE_PIXEL_BLEND(condition, widthvar) \ |
902 | while (condition) { \ |
903 | Uint32 Pixel; \ |
904 | unsigned sR, sG, sB, dR, dG, dB, sA; \ |
905 | DISEMBLE_RGBA(src, 4, srcfmt, Pixel, sR, sG, sB, sA); \ |
906 | if(sA) { \ |
907 | unsigned short dstpixel = *((unsigned short *)dst); \ |
908 | dR = (dstpixel >> 8) & 0xf8; \ |
909 | dG = (dstpixel >> 3) & 0xfc; \ |
910 | dB = (dstpixel << 3) & 0xf8; \ |
911 | ALPHA_BLEND(sR, sG, sB, sA, dR, dG, dB); \ |
912 | *((unsigned short *)dst) = ( \ |
913 | ((dR & 0xf8) << 8) | ((dG & 0xfc) << 3) | (dB >> 3) \ |
914 | ); \ |
915 | } \ |
916 | src += 4; \ |
917 | dst += 2; \ |
918 | widthvar--; \ |
919 | } |
920 | ONE_PIXEL_BLEND((UNALIGNED_PTR(dst)) && (width), width); |
921 | extrawidth = (width % 8); |
922 | valigner = VEC_ALIGNER(src); |
923 | vsrc = (vector unsigned char)vec_ld(0, src); |
924 | width -= extrawidth; |
925 | while (width) { |
926 | vector unsigned char valpha; |
927 | vector unsigned char vsrc1, vsrc2; |
928 | vector unsigned char vdst1, vdst2; |
929 | vector unsigned short vR, vG, vB; |
930 | vector unsigned short vpixel, vrpixel, vgpixel, vbpixel; |
931 | |
932 | /* Load 8 pixels from src as ARGB */ |
933 | voverflow = (vector unsigned char)vec_ld(15, src); |
934 | vsrc = vec_perm(vsrc, voverflow, valigner); |
935 | vsrc1 = vec_perm(vsrc, vsrc, vpermute); |
936 | src += 16; |
937 | vsrc = (vector unsigned char)vec_ld(15, src); |
938 | voverflow = vec_perm(voverflow, vsrc, valigner); |
939 | vsrc2 = vec_perm(voverflow, voverflow, vpermute); |
940 | src += 16; |
941 | |
942 | /* Load 8 pixels from dst as XRGB */ |
943 | voverflow = vec_ld(0, dst); |
944 | vR = vec_and((vector unsigned short)voverflow, vf800); |
945 | vB = vec_sl((vector unsigned short)voverflow, v3_16); |
946 | vG = vec_sl(vB, v2_16); |
947 | vdst1 = (vector unsigned char)vec_perm((vector unsigned char)vR, (vector unsigned char)vR, vredalpha1); |
948 | vdst1 = vec_perm(vdst1, (vector unsigned char)vB, vblue1); |
949 | vdst1 = vec_perm(vdst1, (vector unsigned char)vG, vgreen1); |
950 | vdst2 = (vector unsigned char)vec_perm((vector unsigned char)vR, (vector unsigned char)vR, vredalpha2); |
951 | vdst2 = vec_perm(vdst2, (vector unsigned char)vB, vblue2); |
952 | vdst2 = vec_perm(vdst2, (vector unsigned char)vG, vgreen2); |
953 | |
954 | /* Alpha blend 8 pixels as ARGB */ |
955 | valpha = vec_perm(vsrc1, v0, valphaPermute); |
956 | VEC_MULTIPLY_ALPHA(vsrc1, vdst1, valpha, mergePermute, v1_16, v8_16); |
957 | valpha = vec_perm(vsrc2, v0, valphaPermute); |
958 | VEC_MULTIPLY_ALPHA(vsrc2, vdst2, valpha, mergePermute, v1_16, v8_16); |
959 | |
960 | /* Convert 8 pixels to 565 */ |
961 | vpixel = (vector unsigned short)vec_packpx((vector unsigned int)vdst1, (vector unsigned int)vdst2); |
962 | vgpixel = (vector unsigned short)vec_perm(vdst1, vdst2, vgmerge); |
963 | vgpixel = vec_and(vgpixel, vfc); |
964 | vgpixel = vec_sl(vgpixel, v3_16); |
965 | vrpixel = vec_sl(vpixel, v1_16); |
966 | vrpixel = vec_and(vrpixel, vf800); |
967 | vbpixel = vec_and(vpixel, v3f); |
968 | vdst1 = vec_or((vector unsigned char)vrpixel, (vector unsigned char)vgpixel); |
969 | vdst1 = vec_or(vdst1, (vector unsigned char)vbpixel); |
970 | |
971 | /* Store 8 pixels */ |
972 | vec_st(vdst1, 0, dst); |
973 | |
974 | width -= 8; |
975 | dst += 16; |
976 | } |
977 | ONE_PIXEL_BLEND((extrawidth), extrawidth); |
978 | #undef ONE_PIXEL_BLEND |
979 | src += srcskip; |
980 | dst += dstskip; |
981 | } |
982 | } |
983 | |
984 | static void Blit32to32SurfaceAlphaKeyAltivec(SDL_BlitInfo *info) |
985 | { |
986 | unsigned alpha = info->src->alpha; |
987 | int height = info->d_height; |
988 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
989 | int srcskip = info->s_skip >> 2; |
990 | Uint32 *dstp = (Uint32 *)info->d_pixels; |
991 | int dstskip = info->d_skip >> 2; |
992 | SDL_PixelFormat *srcfmt = info->src; |
993 | SDL_PixelFormat *dstfmt = info->dst; |
994 | unsigned sA = srcfmt->alpha; |
995 | unsigned dA = dstfmt->Amask ? SDL_ALPHA_OPAQUE : 0; |
996 | Uint32 rgbmask = srcfmt->Rmask | srcfmt->Gmask | srcfmt->Bmask; |
997 | Uint32 ckey = info->src->colorkey; |
998 | vector unsigned char mergePermute; |
999 | vector unsigned char vsrcPermute; |
1000 | vector unsigned char vdstPermute; |
1001 | vector unsigned char vsdstPermute; |
1002 | vector unsigned char valpha; |
1003 | vector unsigned char valphamask; |
1004 | vector unsigned char vbits; |
1005 | vector unsigned char v0; |
1006 | vector unsigned short v1; |
1007 | vector unsigned short v8; |
1008 | vector unsigned int vckey; |
1009 | vector unsigned int vrgbmask; |
1010 | |
1011 | mergePermute = VEC_MERGE_PERMUTE(); |
1012 | v0 = vec_splat_u8(0); |
1013 | v1 = vec_splat_u16(1); |
1014 | v8 = vec_splat_u16(8); |
1015 | |
1016 | /* set the alpha to 255 on the destination surf */ |
1017 | valphamask = VEC_ALPHA_MASK(); |
1018 | |
1019 | vsrcPermute = calc_swizzle32(srcfmt, NULL); |
1020 | vdstPermute = calc_swizzle32(NULL, dstfmt); |
1021 | vsdstPermute = calc_swizzle32(dstfmt, NULL); |
1022 | |
1023 | /* set a vector full of alpha and 255-alpha */ |
1024 | ((unsigned char *)&valpha)[0] = alpha; |
1025 | valpha = vec_splat(valpha, 0); |
1026 | vbits = (vector unsigned char)vec_splat_s8(-1); |
1027 | |
1028 | ckey &= rgbmask; |
1029 | ((unsigned int *)(char*)&vckey)[0] = ckey; |
1030 | vckey = vec_splat(vckey, 0); |
1031 | ((unsigned int *)(char*)&vrgbmask)[0] = rgbmask; |
1032 | vrgbmask = vec_splat(vrgbmask, 0); |
1033 | |
1034 | while(height--) { |
1035 | int width = info->d_width; |
1036 | #define ONE_PIXEL_BLEND(condition, widthvar) \ |
1037 | while (condition) { \ |
1038 | Uint32 Pixel; \ |
1039 | unsigned sR, sG, sB, dR, dG, dB; \ |
1040 | RETRIEVE_RGB_PIXEL(((Uint8 *)srcp), 4, Pixel); \ |
1041 | if(sA && Pixel != ckey) { \ |
1042 | RGB_FROM_PIXEL(Pixel, srcfmt, sR, sG, sB); \ |
1043 | DISEMBLE_RGB(((Uint8 *)dstp), 4, dstfmt, Pixel, dR, dG, dB); \ |
1044 | ALPHA_BLEND(sR, sG, sB, sA, dR, dG, dB); \ |
1045 | ASSEMBLE_RGBA(((Uint8 *)dstp), 4, dstfmt, dR, dG, dB, dA); \ |
1046 | } \ |
1047 | dstp++; \ |
1048 | srcp++; \ |
1049 | widthvar--; \ |
1050 | } |
1051 | ONE_PIXEL_BLEND((UNALIGNED_PTR(dstp)) && (width), width); |
1052 | if (width > 0) { |
1053 | int extrawidth = (width % 4); |
1054 | vector unsigned char valigner = VEC_ALIGNER(srcp); |
1055 | vector unsigned char vs = (vector unsigned char)vec_ld(0, srcp); |
1056 | width -= extrawidth; |
1057 | while (width) { |
1058 | vector unsigned char vsel; |
1059 | vector unsigned char voverflow; |
1060 | vector unsigned char vd; |
1061 | vector unsigned char vd_orig; |
1062 | |
1063 | /* s = *srcp */ |
1064 | voverflow = (vector unsigned char)vec_ld(15, srcp); |
1065 | vs = vec_perm(vs, voverflow, valigner); |
1066 | |
1067 | /* vsel is set for items that match the key */ |
1068 | vsel = (vector unsigned char)vec_and((vector unsigned int)vs, vrgbmask); |
1069 | vsel = (vector unsigned char)vec_cmpeq((vector unsigned int)vsel, vckey); |
1070 | |
1071 | /* permute to source format */ |
1072 | vs = vec_perm(vs, valpha, vsrcPermute); |
1073 | |
1074 | /* d = *dstp */ |
1075 | vd = (vector unsigned char)vec_ld(0, dstp); |
1076 | vd_orig = vd = vec_perm(vd, v0, vsdstPermute); |
1077 | |
1078 | VEC_MULTIPLY_ALPHA(vs, vd, valpha, mergePermute, v1, v8); |
1079 | |
1080 | /* set the alpha channel to full on */ |
1081 | vd = vec_or(vd, valphamask); |
1082 | |
1083 | /* mask out color key */ |
1084 | vd = vec_sel(vd, vd_orig, vsel); |
1085 | |
1086 | /* permute to dest format */ |
1087 | vd = vec_perm(vd, vbits, vdstPermute); |
1088 | |
1089 | /* *dstp = res */ |
1090 | vec_st((vector unsigned int)vd, 0, dstp); |
1091 | |
1092 | srcp += 4; |
1093 | dstp += 4; |
1094 | width -= 4; |
1095 | vs = voverflow; |
1096 | } |
1097 | ONE_PIXEL_BLEND((extrawidth), extrawidth); |
1098 | } |
1099 | #undef ONE_PIXEL_BLEND |
1100 | |
1101 | srcp += srcskip; |
1102 | dstp += dstskip; |
1103 | } |
1104 | } |
1105 | |
1106 | |
1107 | static void Blit32to32PixelAlphaAltivec(SDL_BlitInfo *info) |
1108 | { |
1109 | int width = info->d_width; |
1110 | int height = info->d_height; |
1111 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
1112 | int srcskip = info->s_skip >> 2; |
1113 | Uint32 *dstp = (Uint32 *)info->d_pixels; |
1114 | int dstskip = info->d_skip >> 2; |
1115 | SDL_PixelFormat *srcfmt = info->src; |
1116 | SDL_PixelFormat *dstfmt = info->dst; |
1117 | vector unsigned char mergePermute; |
1118 | vector unsigned char valphaPermute; |
1119 | vector unsigned char vsrcPermute; |
1120 | vector unsigned char vdstPermute; |
1121 | vector unsigned char vsdstPermute; |
1122 | vector unsigned char valphamask; |
1123 | vector unsigned char vpixelmask; |
1124 | vector unsigned char v0; |
1125 | vector unsigned short v1; |
1126 | vector unsigned short v8; |
1127 | |
1128 | v0 = vec_splat_u8(0); |
1129 | v1 = vec_splat_u16(1); |
1130 | v8 = vec_splat_u16(8); |
1131 | mergePermute = VEC_MERGE_PERMUTE(); |
1132 | valphamask = VEC_ALPHA_MASK(); |
1133 | valphaPermute = vec_and(vec_lvsl(0, (int *)NULL), vec_splat_u8(0xC)); |
1134 | vpixelmask = vec_nor(valphamask, v0); |
1135 | vsrcPermute = calc_swizzle32(srcfmt, NULL); |
1136 | vdstPermute = calc_swizzle32(NULL, dstfmt); |
1137 | vsdstPermute = calc_swizzle32(dstfmt, NULL); |
1138 | |
1139 | while ( height-- ) { |
1140 | width = info->d_width; |
1141 | #define ONE_PIXEL_BLEND(condition, widthvar) while ((condition)) { \ |
1142 | Uint32 Pixel; \ |
1143 | unsigned sR, sG, sB, dR, dG, dB, sA, dA; \ |
1144 | DISEMBLE_RGBA((Uint8 *)srcp, 4, srcfmt, Pixel, sR, sG, sB, sA); \ |
1145 | if(sA) { \ |
1146 | DISEMBLE_RGBA((Uint8 *)dstp, 4, dstfmt, Pixel, dR, dG, dB, dA); \ |
1147 | ALPHA_BLEND(sR, sG, sB, sA, dR, dG, dB); \ |
1148 | ASSEMBLE_RGBA((Uint8 *)dstp, 4, dstfmt, dR, dG, dB, dA); \ |
1149 | } \ |
1150 | ++srcp; \ |
1151 | ++dstp; \ |
1152 | widthvar--; \ |
1153 | } |
1154 | ONE_PIXEL_BLEND((UNALIGNED_PTR(dstp)) && (width), width); |
1155 | if (width > 0) { |
1156 | /* vsrcPermute */ |
1157 | /* vdstPermute */ |
1158 | int extrawidth = (width % 4); |
1159 | vector unsigned char valigner = VEC_ALIGNER(srcp); |
1160 | vector unsigned char vs = (vector unsigned char)vec_ld(0, srcp); |
1161 | width -= extrawidth; |
1162 | while (width) { |
1163 | vector unsigned char voverflow; |
1164 | vector unsigned char vd; |
1165 | vector unsigned char valpha; |
1166 | vector unsigned char vdstalpha; |
1167 | /* s = *srcp */ |
1168 | voverflow = (vector unsigned char)vec_ld(15, srcp); |
1169 | vs = vec_perm(vs, voverflow, valigner); |
1170 | vs = vec_perm(vs, v0, vsrcPermute); |
1171 | |
1172 | valpha = vec_perm(vs, v0, valphaPermute); |
1173 | |
1174 | /* d = *dstp */ |
1175 | vd = (vector unsigned char)vec_ld(0, dstp); |
1176 | vd = vec_perm(vd, v0, vsdstPermute); |
1177 | vdstalpha = vec_and(vd, valphamask); |
1178 | |
1179 | VEC_MULTIPLY_ALPHA(vs, vd, valpha, mergePermute, v1, v8); |
1180 | |
1181 | /* set the alpha to the dest alpha */ |
1182 | vd = vec_and(vd, vpixelmask); |
1183 | vd = vec_or(vd, vdstalpha); |
1184 | vd = vec_perm(vd, v0, vdstPermute); |
1185 | |
1186 | /* *dstp = res */ |
1187 | vec_st((vector unsigned int)vd, 0, dstp); |
1188 | |
1189 | srcp += 4; |
1190 | dstp += 4; |
1191 | width -= 4; |
1192 | vs = voverflow; |
1193 | |
1194 | } |
1195 | ONE_PIXEL_BLEND((extrawidth), extrawidth); |
1196 | } |
1197 | srcp += srcskip; |
1198 | dstp += dstskip; |
1199 | #undef ONE_PIXEL_BLEND |
1200 | } |
1201 | } |
1202 | |
1203 | /* fast ARGB888->(A)RGB888 blending with pixel alpha */ |
1204 | static void BlitRGBtoRGBPixelAlphaAltivec(SDL_BlitInfo *info) |
1205 | { |
1206 | int width = info->d_width; |
1207 | int height = info->d_height; |
1208 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
1209 | int srcskip = info->s_skip >> 2; |
1210 | Uint32 *dstp = (Uint32 *)info->d_pixels; |
1211 | int dstskip = info->d_skip >> 2; |
1212 | vector unsigned char mergePermute; |
1213 | vector unsigned char valphaPermute; |
1214 | vector unsigned char valphamask; |
1215 | vector unsigned char vpixelmask; |
1216 | vector unsigned char v0; |
1217 | vector unsigned short v1; |
1218 | vector unsigned short v8; |
1219 | v0 = vec_splat_u8(0); |
1220 | v1 = vec_splat_u16(1); |
1221 | v8 = vec_splat_u16(8); |
1222 | mergePermute = VEC_MERGE_PERMUTE(); |
1223 | valphamask = VEC_ALPHA_MASK(); |
1224 | valphaPermute = vec_and(vec_lvsl(0, (int *)NULL), vec_splat_u8(0xC)); |
1225 | |
1226 | |
1227 | vpixelmask = vec_nor(valphamask, v0); |
1228 | while(height--) { |
1229 | width = info->d_width; |
1230 | #define ONE_PIXEL_BLEND(condition, widthvar) \ |
1231 | while ((condition)) { \ |
1232 | Uint32 dalpha; \ |
1233 | Uint32 d; \ |
1234 | Uint32 s1; \ |
1235 | Uint32 d1; \ |
1236 | Uint32 s = *srcp; \ |
1237 | Uint32 alpha = s >> 24; \ |
1238 | if(alpha) { \ |
1239 | if(alpha == SDL_ALPHA_OPAQUE) { \ |
1240 | *dstp = (s & 0x00ffffff) | (*dstp & 0xff000000); \ |
1241 | } else { \ |
1242 | d = *dstp; \ |
1243 | dalpha = d & 0xff000000; \ |
1244 | s1 = s & 0xff00ff; \ |
1245 | d1 = d & 0xff00ff; \ |
1246 | d1 = (d1 + ((s1 - d1) * alpha >> 8)) & 0xff00ff; \ |
1247 | s &= 0xff00; \ |
1248 | d &= 0xff00; \ |
1249 | d = (d + ((s - d) * alpha >> 8)) & 0xff00; \ |
1250 | *dstp = d1 | d | dalpha; \ |
1251 | } \ |
1252 | } \ |
1253 | ++srcp; \ |
1254 | ++dstp; \ |
1255 | widthvar--; \ |
1256 | } |
1257 | ONE_PIXEL_BLEND((UNALIGNED_PTR(dstp)) && (width), width); |
1258 | if (width > 0) { |
1259 | int extrawidth = (width % 4); |
1260 | vector unsigned char valigner = VEC_ALIGNER(srcp); |
1261 | vector unsigned char vs = (vector unsigned char)vec_ld(0, srcp); |
1262 | width -= extrawidth; |
1263 | while (width) { |
1264 | vector unsigned char voverflow; |
1265 | vector unsigned char vd; |
1266 | vector unsigned char valpha; |
1267 | vector unsigned char vdstalpha; |
1268 | /* s = *srcp */ |
1269 | voverflow = (vector unsigned char)vec_ld(15, srcp); |
1270 | vs = vec_perm(vs, voverflow, valigner); |
1271 | |
1272 | valpha = vec_perm(vs, v0, valphaPermute); |
1273 | |
1274 | /* d = *dstp */ |
1275 | vd = (vector unsigned char)vec_ld(0, dstp); |
1276 | vdstalpha = vec_and(vd, valphamask); |
1277 | |
1278 | VEC_MULTIPLY_ALPHA(vs, vd, valpha, mergePermute, v1, v8); |
1279 | |
1280 | /* set the alpha to the dest alpha */ |
1281 | vd = vec_and(vd, vpixelmask); |
1282 | vd = vec_or(vd, vdstalpha); |
1283 | |
1284 | /* *dstp = res */ |
1285 | vec_st((vector unsigned int)vd, 0, dstp); |
1286 | |
1287 | srcp += 4; |
1288 | dstp += 4; |
1289 | width -= 4; |
1290 | vs = voverflow; |
1291 | } |
1292 | ONE_PIXEL_BLEND((extrawidth), extrawidth); |
1293 | } |
1294 | srcp += srcskip; |
1295 | dstp += dstskip; |
1296 | } |
1297 | #undef ONE_PIXEL_BLEND |
1298 | } |
1299 | |
1300 | static void Blit32to32SurfaceAlphaAltivec(SDL_BlitInfo *info) |
1301 | { |
1302 | /* XXX : 6 */ |
1303 | unsigned alpha = info->src->alpha; |
1304 | int height = info->d_height; |
1305 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
1306 | int srcskip = info->s_skip >> 2; |
1307 | Uint32 *dstp = (Uint32 *)info->d_pixels; |
1308 | int dstskip = info->d_skip >> 2; |
1309 | SDL_PixelFormat *srcfmt = info->src; |
1310 | SDL_PixelFormat *dstfmt = info->dst; |
1311 | unsigned sA = srcfmt->alpha; |
1312 | unsigned dA = dstfmt->Amask ? SDL_ALPHA_OPAQUE : 0; |
1313 | vector unsigned char mergePermute; |
1314 | vector unsigned char vsrcPermute; |
1315 | vector unsigned char vdstPermute; |
1316 | vector unsigned char vsdstPermute; |
1317 | vector unsigned char valpha; |
1318 | vector unsigned char valphamask; |
1319 | vector unsigned char vbits; |
1320 | vector unsigned short v1; |
1321 | vector unsigned short v8; |
1322 | |
1323 | mergePermute = VEC_MERGE_PERMUTE(); |
1324 | v1 = vec_splat_u16(1); |
1325 | v8 = vec_splat_u16(8); |
1326 | |
1327 | /* set the alpha to 255 on the destination surf */ |
1328 | valphamask = VEC_ALPHA_MASK(); |
1329 | |
1330 | vsrcPermute = calc_swizzle32(srcfmt, NULL); |
1331 | vdstPermute = calc_swizzle32(NULL, dstfmt); |
1332 | vsdstPermute = calc_swizzle32(dstfmt, NULL); |
1333 | |
1334 | /* set a vector full of alpha and 255-alpha */ |
1335 | ((unsigned char *)&valpha)[0] = alpha; |
1336 | valpha = vec_splat(valpha, 0); |
1337 | vbits = (vector unsigned char)vec_splat_s8(-1); |
1338 | |
1339 | while(height--) { |
1340 | int width = info->d_width; |
1341 | #define ONE_PIXEL_BLEND(condition, widthvar) while ((condition)) { \ |
1342 | Uint32 Pixel; \ |
1343 | unsigned sR, sG, sB, dR, dG, dB; \ |
1344 | DISEMBLE_RGB(((Uint8 *)srcp), 4, srcfmt, Pixel, sR, sG, sB); \ |
1345 | DISEMBLE_RGB(((Uint8 *)dstp), 4, dstfmt, Pixel, dR, dG, dB); \ |
1346 | ALPHA_BLEND(sR, sG, sB, sA, dR, dG, dB); \ |
1347 | ASSEMBLE_RGBA(((Uint8 *)dstp), 4, dstfmt, dR, dG, dB, dA); \ |
1348 | ++srcp; \ |
1349 | ++dstp; \ |
1350 | widthvar--; \ |
1351 | } |
1352 | ONE_PIXEL_BLEND((UNALIGNED_PTR(dstp)) && (width), width); |
1353 | if (width > 0) { |
1354 | int extrawidth = (width % 4); |
1355 | vector unsigned char valigner = VEC_ALIGNER(srcp); |
1356 | vector unsigned char vs = (vector unsigned char)vec_ld(0, srcp); |
1357 | width -= extrawidth; |
1358 | while (width) { |
1359 | vector unsigned char voverflow; |
1360 | vector unsigned char vd; |
1361 | |
1362 | /* s = *srcp */ |
1363 | voverflow = (vector unsigned char)vec_ld(15, srcp); |
1364 | vs = vec_perm(vs, voverflow, valigner); |
1365 | vs = vec_perm(vs, valpha, vsrcPermute); |
1366 | |
1367 | /* d = *dstp */ |
1368 | vd = (vector unsigned char)vec_ld(0, dstp); |
1369 | vd = vec_perm(vd, vd, vsdstPermute); |
1370 | |
1371 | VEC_MULTIPLY_ALPHA(vs, vd, valpha, mergePermute, v1, v8); |
1372 | |
1373 | /* set the alpha channel to full on */ |
1374 | vd = vec_or(vd, valphamask); |
1375 | vd = vec_perm(vd, vbits, vdstPermute); |
1376 | |
1377 | /* *dstp = res */ |
1378 | vec_st((vector unsigned int)vd, 0, dstp); |
1379 | |
1380 | srcp += 4; |
1381 | dstp += 4; |
1382 | width -= 4; |
1383 | vs = voverflow; |
1384 | } |
1385 | ONE_PIXEL_BLEND((extrawidth), extrawidth); |
1386 | } |
1387 | #undef ONE_PIXEL_BLEND |
1388 | |
1389 | srcp += srcskip; |
1390 | dstp += dstskip; |
1391 | } |
1392 | |
1393 | } |
1394 | |
1395 | |
1396 | /* fast RGB888->(A)RGB888 blending */ |
1397 | static void BlitRGBtoRGBSurfaceAlphaAltivec(SDL_BlitInfo *info) |
1398 | { |
1399 | unsigned alpha = info->src->alpha; |
1400 | int height = info->d_height; |
1401 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
1402 | int srcskip = info->s_skip >> 2; |
1403 | Uint32 *dstp = (Uint32 *)info->d_pixels; |
1404 | int dstskip = info->d_skip >> 2; |
1405 | vector unsigned char mergePermute; |
1406 | vector unsigned char valpha; |
1407 | vector unsigned char valphamask; |
1408 | vector unsigned short v1; |
1409 | vector unsigned short v8; |
1410 | |
1411 | mergePermute = VEC_MERGE_PERMUTE(); |
1412 | v1 = vec_splat_u16(1); |
1413 | v8 = vec_splat_u16(8); |
1414 | |
1415 | /* set the alpha to 255 on the destination surf */ |
1416 | valphamask = VEC_ALPHA_MASK(); |
1417 | |
1418 | /* set a vector full of alpha and 255-alpha */ |
1419 | ((unsigned char *)&valpha)[0] = alpha; |
1420 | valpha = vec_splat(valpha, 0); |
1421 | |
1422 | while(height--) { |
1423 | int width = info->d_width; |
1424 | #define ONE_PIXEL_BLEND(condition, widthvar) while ((condition)) { \ |
1425 | Uint32 s = *srcp; \ |
1426 | Uint32 d = *dstp; \ |
1427 | Uint32 s1 = s & 0xff00ff; \ |
1428 | Uint32 d1 = d & 0xff00ff; \ |
1429 | d1 = (d1 + ((s1 - d1) * alpha >> 8)) \ |
1430 | & 0xff00ff; \ |
1431 | s &= 0xff00; \ |
1432 | d &= 0xff00; \ |
1433 | d = (d + ((s - d) * alpha >> 8)) & 0xff00; \ |
1434 | *dstp = d1 | d | 0xff000000; \ |
1435 | ++srcp; \ |
1436 | ++dstp; \ |
1437 | widthvar--; \ |
1438 | } |
1439 | ONE_PIXEL_BLEND((UNALIGNED_PTR(dstp)) && (width), width); |
1440 | if (width > 0) { |
1441 | int extrawidth = (width % 4); |
1442 | vector unsigned char valigner = VEC_ALIGNER(srcp); |
1443 | vector unsigned char vs = (vector unsigned char)vec_ld(0, srcp); |
1444 | width -= extrawidth; |
1445 | while (width) { |
1446 | vector unsigned char voverflow; |
1447 | vector unsigned char vd; |
1448 | |
1449 | /* s = *srcp */ |
1450 | voverflow = (vector unsigned char)vec_ld(15, srcp); |
1451 | vs = vec_perm(vs, voverflow, valigner); |
1452 | |
1453 | /* d = *dstp */ |
1454 | vd = (vector unsigned char)vec_ld(0, dstp); |
1455 | |
1456 | VEC_MULTIPLY_ALPHA(vs, vd, valpha, mergePermute, v1, v8); |
1457 | |
1458 | /* set the alpha channel to full on */ |
1459 | vd = vec_or(vd, valphamask); |
1460 | |
1461 | /* *dstp = res */ |
1462 | vec_st((vector unsigned int)vd, 0, dstp); |
1463 | |
1464 | srcp += 4; |
1465 | dstp += 4; |
1466 | width -= 4; |
1467 | vs = voverflow; |
1468 | } |
1469 | ONE_PIXEL_BLEND((extrawidth), extrawidth); |
1470 | } |
1471 | #undef ONE_PIXEL_BLEND |
1472 | |
1473 | srcp += srcskip; |
1474 | dstp += dstskip; |
1475 | } |
1476 | } |
1477 | #if __MWERKS__ |
1478 | #pragma altivec_model off |
1479 | #endif |
1480 | #endif /* SDL_ALTIVEC_BLITTERS */ |
1481 | |
1482 | /* fast RGB888->(A)RGB888 blending with surface alpha=128 special case */ |
1483 | static void BlitRGBtoRGBSurfaceAlpha128(SDL_BlitInfo *info) |
1484 | { |
1485 | int width = info->d_width; |
1486 | int height = info->d_height; |
1487 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
1488 | int srcskip = info->s_skip >> 2; |
1489 | Uint32 *dstp = (Uint32 *)info->d_pixels; |
1490 | int dstskip = info->d_skip >> 2; |
1491 | |
1492 | while(height--) { |
1493 | DUFFS_LOOP4({ |
1494 | Uint32 s = *srcp++; |
1495 | Uint32 d = *dstp; |
1496 | *dstp++ = ((((s & 0x00fefefe) + (d & 0x00fefefe)) >> 1) |
1497 | + (s & d & 0x00010101)) | 0xff000000; |
1498 | }, width); |
1499 | srcp += srcskip; |
1500 | dstp += dstskip; |
1501 | } |
1502 | } |
1503 | |
1504 | /* fast RGB888->(A)RGB888 blending with surface alpha */ |
1505 | static void BlitRGBtoRGBSurfaceAlpha(SDL_BlitInfo *info) |
1506 | { |
1507 | unsigned alpha = info->src->alpha; |
1508 | if(alpha == 128) { |
1509 | BlitRGBtoRGBSurfaceAlpha128(info); |
1510 | } else { |
1511 | int width = info->d_width; |
1512 | int height = info->d_height; |
1513 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
1514 | int srcskip = info->s_skip >> 2; |
1515 | Uint32 *dstp = (Uint32 *)info->d_pixels; |
1516 | int dstskip = info->d_skip >> 2; |
1517 | Uint32 s; |
1518 | Uint32 d; |
1519 | Uint32 s1; |
1520 | Uint32 d1; |
1521 | |
1522 | while(height--) { |
1523 | DUFFS_LOOP_DOUBLE2({ |
1524 | /* One Pixel Blend */ |
1525 | s = *srcp; |
1526 | d = *dstp; |
1527 | s1 = s & 0xff00ff; |
1528 | d1 = d & 0xff00ff; |
1529 | d1 = (d1 + ((s1 - d1) * alpha >> 8)) |
1530 | & 0xff00ff; |
1531 | s &= 0xff00; |
1532 | d &= 0xff00; |
1533 | d = (d + ((s - d) * alpha >> 8)) & 0xff00; |
1534 | *dstp = d1 | d | 0xff000000; |
1535 | ++srcp; |
1536 | ++dstp; |
1537 | },{ |
1538 | /* Two Pixels Blend */ |
1539 | s = *srcp; |
1540 | d = *dstp; |
1541 | s1 = s & 0xff00ff; |
1542 | d1 = d & 0xff00ff; |
1543 | d1 += (s1 - d1) * alpha >> 8; |
1544 | d1 &= 0xff00ff; |
1545 | |
1546 | s = ((s & 0xff00) >> 8) | |
1547 | ((srcp[1] & 0xff00) << 8); |
1548 | d = ((d & 0xff00) >> 8) | |
1549 | ((dstp[1] & 0xff00) << 8); |
1550 | d += (s - d) * alpha >> 8; |
1551 | d &= 0x00ff00ff; |
1552 | |
1553 | *dstp++ = d1 | ((d << 8) & 0xff00) | 0xff000000; |
1554 | ++srcp; |
1555 | |
1556 | s1 = *srcp; |
1557 | d1 = *dstp; |
1558 | s1 &= 0xff00ff; |
1559 | d1 &= 0xff00ff; |
1560 | d1 += (s1 - d1) * alpha >> 8; |
1561 | d1 &= 0xff00ff; |
1562 | |
1563 | *dstp = d1 | ((d >> 8) & 0xff00) | 0xff000000; |
1564 | ++srcp; |
1565 | ++dstp; |
1566 | }, width); |
1567 | srcp += srcskip; |
1568 | dstp += dstskip; |
1569 | } |
1570 | } |
1571 | } |
1572 | |
1573 | /* fast ARGB888->(A)RGB888 blending with pixel alpha */ |
1574 | static void BlitRGBtoRGBPixelAlpha(SDL_BlitInfo *info) |
1575 | { |
1576 | int width = info->d_width; |
1577 | int height = info->d_height; |
1578 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
1579 | int srcskip = info->s_skip >> 2; |
1580 | Uint32 *dstp = (Uint32 *)info->d_pixels; |
1581 | int dstskip = info->d_skip >> 2; |
1582 | |
1583 | while(height--) { |
1584 | DUFFS_LOOP4({ |
1585 | Uint32 dalpha; |
1586 | Uint32 d; |
1587 | Uint32 s1; |
1588 | Uint32 d1; |
1589 | Uint32 s = *srcp; |
1590 | Uint32 alpha = s >> 24; |
1591 | /* FIXME: Here we special-case opaque alpha since the |
1592 | compositioning used (>>8 instead of /255) doesn't handle |
1593 | it correctly. Also special-case alpha=0 for speed? |
1594 | Benchmark this! */ |
1595 | if(alpha) { |
1596 | if(alpha == SDL_ALPHA_OPAQUE) { |
1597 | *dstp = (s & 0x00ffffff) | (*dstp & 0xff000000); |
1598 | } else { |
1599 | /* |
1600 | * take out the middle component (green), and process |
1601 | * the other two in parallel. One multiply less. |
1602 | */ |
1603 | d = *dstp; |
1604 | dalpha = d & 0xff000000; |
1605 | s1 = s & 0xff00ff; |
1606 | d1 = d & 0xff00ff; |
1607 | d1 = (d1 + ((s1 - d1) * alpha >> 8)) & 0xff00ff; |
1608 | s &= 0xff00; |
1609 | d &= 0xff00; |
1610 | d = (d + ((s - d) * alpha >> 8)) & 0xff00; |
1611 | *dstp = d1 | d | dalpha; |
1612 | } |
1613 | } |
1614 | ++srcp; |
1615 | ++dstp; |
1616 | }, width); |
1617 | srcp += srcskip; |
1618 | dstp += dstskip; |
1619 | } |
1620 | } |
1621 | |
1622 | #if GCC_ASMBLIT |
1623 | /* fast (as in MMX with prefetch) ARGB888->(A)RGB888 blending with pixel alpha */ |
1624 | static void BlitRGBtoRGBPixelAlphaMMX3DNOW(SDL_BlitInfo *info) |
1625 | { |
1626 | int width = info->d_width; |
1627 | int height = info->d_height; |
1628 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
1629 | int srcskip = info->s_skip >> 2; |
1630 | Uint32 *dstp = (Uint32 *)info->d_pixels; |
1631 | int dstskip = info->d_skip >> 2; |
1632 | SDL_PixelFormat* sf = info->src; |
1633 | Uint32 amask = sf->Amask; |
1634 | |
1635 | __asm__ ( |
1636 | /* make mm6 all zeros. */ |
1637 | "pxor %%mm6, %%mm6\n" |
1638 | |
1639 | /* Make a mask to preserve the alpha. */ |
1640 | "movd %0, %%mm7\n\t" /* 0000F000 -> mm7 */ |
1641 | "punpcklbw %%mm7, %%mm7\n\t" /* FF000000 -> mm7 */ |
1642 | "pcmpeqb %%mm4, %%mm4\n\t" /* FFFFFFFF -> mm4 */ |
1643 | "movq %%mm4, %%mm3\n\t" /* FFFFFFFF -> mm3 (for later) */ |
1644 | "pxor %%mm4, %%mm7\n\t" /* 00FFFFFF -> mm7 (mult mask) */ |
1645 | |
1646 | /* form channel masks */ |
1647 | "movq %%mm7, %%mm4\n\t" /* 00FFFFFF -> mm4 */ |
1648 | "packsswb %%mm6, %%mm4\n\t" /* 00000FFF -> mm4 (channel mask) */ |
1649 | "packsswb %%mm6, %%mm3\n\t" /* 0000FFFF -> mm3 */ |
1650 | "pxor %%mm4, %%mm3\n\t" /* 0000F000 -> mm3 (~channel mask) */ |
1651 | |
1652 | /* get alpha channel shift */ |
1653 | "movd %1, %%mm5\n\t" /* Ashift -> mm5 */ |
1654 | |
1655 | : /* nothing */ : "rm" (amask), "rm" ((Uint32) sf->Ashift) ); |
1656 | |
1657 | while(height--) { |
1658 | |
1659 | DUFFS_LOOP4({ |
1660 | Uint32 alpha; |
1661 | |
1662 | __asm__ ( |
1663 | "prefetch 64(%0)\n" |
1664 | "prefetch 64(%1)\n" |
1665 | : : "r" (srcp), "r" (dstp) ); |
1666 | |
1667 | alpha = *srcp & amask; |
1668 | /* FIXME: Here we special-case opaque alpha since the |
1669 | compositioning used (>>8 instead of /255) doesn't handle |
1670 | it correctly. Also special-case alpha=0 for speed? |
1671 | Benchmark this! */ |
1672 | if(alpha == 0) { |
1673 | /* do nothing */ |
1674 | } |
1675 | else if(alpha == amask) { |
1676 | /* opaque alpha -- copy RGB, keep dst alpha */ |
1677 | /* using MMX here to free up regular registers for other things */ |
1678 | __asm__ ( |
1679 | "movd (%0), %%mm0\n\t" /* src(ARGB) -> mm0 (0000ARGB)*/ |
1680 | "movd (%1), %%mm1\n\t" /* dst(ARGB) -> mm1 (0000ARGB)*/ |
1681 | "pand %%mm4, %%mm0\n\t" /* src & chanmask -> mm0 */ |
1682 | "pand %%mm3, %%mm1\n\t" /* dst & ~chanmask -> mm2 */ |
1683 | "por %%mm0, %%mm1\n\t" /* src | dst -> mm1 */ |
1684 | "movd %%mm1, (%1) \n\t" /* mm1 -> dst */ |
1685 | |
1686 | : : "r" (srcp), "r" (dstp) ); |
1687 | } |
1688 | |
1689 | else { |
1690 | __asm__ ( |
1691 | /* load in the source, and dst. */ |
1692 | "movd (%0), %%mm0\n" /* mm0(s) = 0 0 0 0 | As Rs Gs Bs */ |
1693 | "movd (%1), %%mm1\n" /* mm1(d) = 0 0 0 0 | Ad Rd Gd Bd */ |
1694 | |
1695 | /* Move the src alpha into mm2 */ |
1696 | |
1697 | /* if supporting pshufw */ |
1698 | /*"pshufw $0x55, %%mm0, %%mm2\n" */ /* mm2 = 0 As 0 As | 0 As 0 As */ |
1699 | /*"psrlw $8, %%mm2\n" */ |
1700 | |
1701 | /* else: */ |
1702 | "movd %2, %%mm2\n" |
1703 | "psrld %%mm5, %%mm2\n" /* mm2 = 0 0 0 0 | 0 0 0 As */ |
1704 | "punpcklwd %%mm2, %%mm2\n" /* mm2 = 0 0 0 0 | 0 As 0 As */ |
1705 | "punpckldq %%mm2, %%mm2\n" /* mm2 = 0 As 0 As | 0 As 0 As */ |
1706 | "pand %%mm7, %%mm2\n" /* to preserve dest alpha */ |
1707 | |
1708 | /* move the colors into words. */ |
1709 | "punpcklbw %%mm6, %%mm0\n" /* mm0 = 0 As 0 Rs | 0 Gs 0 Bs */ |
1710 | "punpcklbw %%mm6, %%mm1\n" /* mm0 = 0 Ad 0 Rd | 0 Gd 0 Bd */ |
1711 | |
1712 | /* src - dst */ |
1713 | "psubw %%mm1, %%mm0\n" /* mm0 = As-Ad Rs-Rd | Gs-Gd Bs-Bd */ |
1714 | |
1715 | /* A * (src-dst) */ |
1716 | "pmullw %%mm2, %%mm0\n" /* mm0 = 0*As-d As*Rs-d | As*Gs-d As*Bs-d */ |
1717 | "psrlw $8, %%mm0\n" /* mm0 = 0>>8 Rc>>8 | Gc>>8 Bc>>8 */ |
1718 | "paddb %%mm1, %%mm0\n" /* mm0 = 0+Ad Rc+Rd | Gc+Gd Bc+Bd */ |
1719 | |
1720 | "packuswb %%mm0, %%mm0\n" /* mm0 = | Ac Rc Gc Bc */ |
1721 | |
1722 | "movd %%mm0, (%1)\n" /* result in mm0 */ |
1723 | |
1724 | : : "r" (srcp), "r" (dstp), "r" (alpha) ); |
1725 | |
1726 | } |
1727 | ++srcp; |
1728 | ++dstp; |
1729 | }, width); |
1730 | srcp += srcskip; |
1731 | dstp += dstskip; |
1732 | } |
1733 | |
1734 | __asm__ ( |
1735 | "emms\n" |
1736 | : ); |
1737 | } |
1738 | /* End GCC_ASMBLIT*/ |
1739 | |
1740 | #elif MSVC_ASMBLIT |
1741 | /* fast (as in MMX with prefetch) ARGB888->(A)RGB888 blending with pixel alpha */ |
1742 | static void BlitRGBtoRGBPixelAlphaMMX3DNOW(SDL_BlitInfo *info) |
1743 | { |
1744 | int width = info->d_width; |
1745 | int height = info->d_height; |
1746 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
1747 | int srcskip = info->s_skip >> 2; |
1748 | Uint32 *dstp = (Uint32 *)info->d_pixels; |
1749 | int dstskip = info->d_skip >> 2; |
1750 | SDL_PixelFormat* sf = info->src; |
1751 | Uint32 chanmask = sf->Rmask | sf->Gmask | sf->Bmask; |
1752 | Uint32 amask = sf->Amask; |
1753 | Uint32 ashift = sf->Ashift; |
1754 | Uint64 multmask; |
1755 | |
1756 | __m64 src1, dst1, mm_alpha, mm_zero, dmask; |
1757 | |
1758 | mm_zero = _mm_setzero_si64(); /* 0 -> mm_zero */ |
1759 | multmask = ~(0xFFFFi64 << (ashift * 2)); |
1760 | dmask = *(__m64*) &multmask; /* dst alpha mask -> dmask */ |
1761 | |
1762 | while(height--) { |
1763 | DUFFS_LOOP4({ |
1764 | Uint32 alpha; |
1765 | |
1766 | _m_prefetch(srcp + 16); |
1767 | _m_prefetch(dstp + 16); |
1768 | |
1769 | alpha = *srcp & amask; |
1770 | if (alpha == 0) { |
1771 | /* do nothing */ |
1772 | } else if (alpha == amask) { |
1773 | /* copy RGB, keep dst alpha */ |
1774 | *dstp = (*srcp & chanmask) | (*dstp & ~chanmask); |
1775 | } else { |
1776 | src1 = _mm_cvtsi32_si64(*srcp); /* src(ARGB) -> src1 (0000ARGB)*/ |
1777 | src1 = _mm_unpacklo_pi8(src1, mm_zero); /* 0A0R0G0B -> src1 */ |
1778 | |
1779 | dst1 = _mm_cvtsi32_si64(*dstp); /* dst(ARGB) -> dst1 (0000ARGB)*/ |
1780 | dst1 = _mm_unpacklo_pi8(dst1, mm_zero); /* 0A0R0G0B -> dst1 */ |
1781 | |
1782 | mm_alpha = _mm_cvtsi32_si64(alpha); /* alpha -> mm_alpha (0000000A) */ |
1783 | mm_alpha = _mm_srli_si64(mm_alpha, ashift); /* mm_alpha >> ashift -> mm_alpha(0000000A) */ |
1784 | mm_alpha = _mm_unpacklo_pi16(mm_alpha, mm_alpha); /* 00000A0A -> mm_alpha */ |
1785 | mm_alpha = _mm_unpacklo_pi32(mm_alpha, mm_alpha); /* 0A0A0A0A -> mm_alpha */ |
1786 | mm_alpha = _mm_and_si64(mm_alpha, dmask); /* 000A0A0A -> mm_alpha, preserve dst alpha on add */ |
1787 | |
1788 | /* blend */ |
1789 | src1 = _mm_sub_pi16(src1, dst1);/* src - dst -> src1 */ |
1790 | src1 = _mm_mullo_pi16(src1, mm_alpha); /* (src - dst) * alpha -> src1 */ |
1791 | src1 = _mm_srli_pi16(src1, 8); /* src1 >> 8 -> src1(000R0G0B) */ |
1792 | dst1 = _mm_add_pi8(src1, dst1); /* src1 + dst1(dst) -> dst1(0A0R0G0B) */ |
1793 | dst1 = _mm_packs_pu16(dst1, mm_zero); /* 0000ARGB -> dst1 */ |
1794 | |
1795 | *dstp = _mm_cvtsi64_si32(dst1); /* dst1 -> pixel */ |
1796 | } |
1797 | ++srcp; |
1798 | ++dstp; |
1799 | }, width); |
1800 | srcp += srcskip; |
1801 | dstp += dstskip; |
1802 | } |
1803 | _mm_empty(); |
1804 | } |
1805 | /* End MSVC_ASMBLIT */ |
1806 | |
1807 | #endif /* GCC_ASMBLIT, MSVC_ASMBLIT */ |
1808 | |
1809 | /* 16bpp special case for per-surface alpha=50%: blend 2 pixels in parallel */ |
1810 | |
1811 | /* blend a single 16 bit pixel at 50% */ |
1812 | #define BLEND16_50(d, s, mask) \ |
1813 | ((((s & mask) + (d & mask)) >> 1) + (s & d & (~mask & 0xffff))) |
1814 | |
1815 | /* blend two 16 bit pixels at 50% */ |
1816 | #define BLEND2x16_50(d, s, mask) \ |
1817 | (((s & (mask | mask << 16)) >> 1) + ((d & (mask | mask << 16)) >> 1) \ |
1818 | + (s & d & (~(mask | mask << 16)))) |
1819 | |
1820 | static void Blit16to16SurfaceAlpha128(SDL_BlitInfo *info, Uint16 mask) |
1821 | { |
1822 | int width = info->d_width; |
1823 | int height = info->d_height; |
1824 | Uint16 *srcp = (Uint16 *)info->s_pixels; |
1825 | int srcskip = info->s_skip >> 1; |
1826 | Uint16 *dstp = (Uint16 *)info->d_pixels; |
1827 | int dstskip = info->d_skip >> 1; |
1828 | |
1829 | while(height--) { |
1830 | if(((uintptr_t)srcp ^ (uintptr_t)dstp) & 2) { |
1831 | /* |
1832 | * Source and destination not aligned, pipeline it. |
1833 | * This is mostly a win for big blits but no loss for |
1834 | * small ones |
1835 | */ |
1836 | Uint32 prev_sw; |
1837 | int w = width; |
1838 | |
1839 | /* handle odd destination */ |
1840 | if((uintptr_t)dstp & 2) { |
1841 | Uint16 d = *dstp, s = *srcp; |
1842 | *dstp = BLEND16_50(d, s, mask); |
1843 | dstp++; |
1844 | srcp++; |
1845 | w--; |
1846 | } |
1847 | srcp++; /* srcp is now 32-bit aligned */ |
1848 | |
1849 | /* bootstrap pipeline with first halfword */ |
1850 | prev_sw = ((Uint32 *)srcp)[-1]; |
1851 | |
1852 | while(w > 1) { |
1853 | Uint32 sw, dw, s; |
1854 | sw = *(Uint32 *)srcp; |
1855 | dw = *(Uint32 *)dstp; |
1856 | #if SDL_BYTEORDER == SDL_BIG_ENDIAN |
1857 | s = (prev_sw << 16) + (sw >> 16); |
1858 | #else |
1859 | s = (prev_sw >> 16) + (sw << 16); |
1860 | #endif |
1861 | prev_sw = sw; |
1862 | *(Uint32 *)dstp = BLEND2x16_50(dw, s, mask); |
1863 | dstp += 2; |
1864 | srcp += 2; |
1865 | w -= 2; |
1866 | } |
1867 | |
1868 | /* final pixel if any */ |
1869 | if(w) { |
1870 | Uint16 d = *dstp, s; |
1871 | #if SDL_BYTEORDER == SDL_BIG_ENDIAN |
1872 | s = (Uint16)prev_sw; |
1873 | #else |
1874 | s = (Uint16)(prev_sw >> 16); |
1875 | #endif |
1876 | *dstp = BLEND16_50(d, s, mask); |
1877 | srcp++; |
1878 | dstp++; |
1879 | } |
1880 | srcp += srcskip - 1; |
1881 | dstp += dstskip; |
1882 | } else { |
1883 | /* source and destination are aligned */ |
1884 | int w = width; |
1885 | |
1886 | /* first odd pixel? */ |
1887 | if((uintptr_t)srcp & 2) { |
1888 | Uint16 d = *dstp, s = *srcp; |
1889 | *dstp = BLEND16_50(d, s, mask); |
1890 | srcp++; |
1891 | dstp++; |
1892 | w--; |
1893 | } |
1894 | /* srcp and dstp are now 32-bit aligned */ |
1895 | |
1896 | while(w > 1) { |
1897 | Uint32 sw = *(Uint32 *)srcp; |
1898 | Uint32 dw = *(Uint32 *)dstp; |
1899 | *(Uint32 *)dstp = BLEND2x16_50(dw, sw, mask); |
1900 | srcp += 2; |
1901 | dstp += 2; |
1902 | w -= 2; |
1903 | } |
1904 | |
1905 | /* last odd pixel? */ |
1906 | if(w) { |
1907 | Uint16 d = *dstp, s = *srcp; |
1908 | *dstp = BLEND16_50(d, s, mask); |
1909 | srcp++; |
1910 | dstp++; |
1911 | } |
1912 | srcp += srcskip; |
1913 | dstp += dstskip; |
1914 | } |
1915 | } |
1916 | } |
1917 | |
1918 | #if GCC_ASMBLIT |
1919 | /* fast RGB565->RGB565 blending with surface alpha */ |
1920 | static void Blit565to565SurfaceAlphaMMX(SDL_BlitInfo *info) |
1921 | { |
1922 | unsigned alpha = info->src->alpha; /* downscale alpha to 5 bits */ |
1923 | if(alpha == 128) { |
1924 | Blit16to16SurfaceAlpha128(info, 0xf7de); |
1925 | } else { |
1926 | int width = info->d_width; |
1927 | int height = info->d_height; |
1928 | Uint16 *srcp = (Uint16 *)info->s_pixels; |
1929 | int srcskip = info->s_skip >> 1; |
1930 | Uint16 *dstp = (Uint16 *)info->d_pixels; |
1931 | int dstskip = info->d_skip >> 1; |
1932 | Uint32 s, d; |
1933 | Uint64 load; |
1934 | |
1935 | alpha &= ~(1+2+4); /* cut alpha to get the exact same behaviour */ |
1936 | load = alpha; |
1937 | alpha >>= 3; /* downscale alpha to 5 bits */ |
1938 | |
1939 | movq_m2r(load, mm0); /* alpha(0000000A) -> mm0 */ |
1940 | punpcklwd_r2r(mm0, mm0); /* 00000A0A -> mm0 */ |
1941 | punpcklwd_r2r(mm0, mm0); /* 0A0A0A0A -> mm0 */ |
1942 | /* position alpha to allow for mullo and mulhi on diff channels |
1943 | to reduce the number of operations */ |
1944 | psllq_i2r(3, mm0); |
1945 | |
1946 | /* Setup the 565 color channel masks */ |
1947 | load = 0x07E007E007E007E0ULL; |
1948 | movq_m2r(load, mm4); /* MASKGREEN -> mm4 */ |
1949 | load = 0x001F001F001F001FULL; |
1950 | movq_m2r(load, mm7); /* MASKBLUE -> mm7 */ |
1951 | while(height--) { |
1952 | DUFFS_LOOP_QUATRO2( |
1953 | { |
1954 | s = *srcp++; |
1955 | d = *dstp; |
1956 | /* |
1957 | * shift out the middle component (green) to |
1958 | * the high 16 bits, and process all three RGB |
1959 | * components at the same time. |
1960 | */ |
1961 | s = (s | s << 16) & 0x07e0f81f; |
1962 | d = (d | d << 16) & 0x07e0f81f; |
1963 | d += (s - d) * alpha >> 5; |
1964 | d &= 0x07e0f81f; |
1965 | *dstp++ = d | d >> 16; |
1966 | },{ |
1967 | s = *srcp++; |
1968 | d = *dstp; |
1969 | /* |
1970 | * shift out the middle component (green) to |
1971 | * the high 16 bits, and process all three RGB |
1972 | * components at the same time. |
1973 | */ |
1974 | s = (s | s << 16) & 0x07e0f81f; |
1975 | d = (d | d << 16) & 0x07e0f81f; |
1976 | d += (s - d) * alpha >> 5; |
1977 | d &= 0x07e0f81f; |
1978 | *dstp++ = d | d >> 16; |
1979 | s = *srcp++; |
1980 | d = *dstp; |
1981 | /* |
1982 | * shift out the middle component (green) to |
1983 | * the high 16 bits, and process all three RGB |
1984 | * components at the same time. |
1985 | */ |
1986 | s = (s | s << 16) & 0x07e0f81f; |
1987 | d = (d | d << 16) & 0x07e0f81f; |
1988 | d += (s - d) * alpha >> 5; |
1989 | d &= 0x07e0f81f; |
1990 | *dstp++ = d | d >> 16; |
1991 | },{ |
1992 | movq_m2r((*srcp), mm2);/* 4 src pixels -> mm2 */ |
1993 | movq_m2r((*dstp), mm3);/* 4 dst pixels -> mm3 */ |
1994 | |
1995 | /* red -- does not need a mask since the right shift clears |
1996 | the uninteresting bits */ |
1997 | movq_r2r(mm2, mm5); /* src -> mm5 */ |
1998 | movq_r2r(mm3, mm6); /* dst -> mm6 */ |
1999 | psrlw_i2r(11, mm5); /* mm5 >> 11 -> mm5 [000r 000r 000r 000r] */ |
2000 | psrlw_i2r(11, mm6); /* mm6 >> 11 -> mm6 [000r 000r 000r 000r] */ |
2001 | |
2002 | /* blend */ |
2003 | psubw_r2r(mm6, mm5);/* src - dst -> mm5 */ |
2004 | pmullw_r2r(mm0, mm5); /* mm5 * alpha -> mm5 */ |
2005 | /* alpha used is actually 11 bits |
2006 | 11 + 5 = 16 bits, so the sign bits are lost */ |
2007 | psrlw_i2r(11, mm5); /* mm5 >> 11 -> mm5 */ |
2008 | paddw_r2r(mm5, mm6); /* mm5 + mm6(dst) -> mm6 */ |
2009 | psllw_i2r(11, mm6); /* mm6 << 11 -> mm6 */ |
2010 | |
2011 | movq_r2r(mm6, mm1); /* save new reds in dsts */ |
2012 | |
2013 | /* green -- process the bits in place */ |
2014 | movq_r2r(mm2, mm5); /* src -> mm5 */ |
2015 | movq_r2r(mm3, mm6); /* dst -> mm6 */ |
2016 | pand_r2r(mm4, mm5); /* src & MASKGREEN -> mm5 */ |
2017 | pand_r2r(mm4, mm6); /* dst & MASKGREEN -> mm6 */ |
2018 | |
2019 | /* blend */ |
2020 | psubw_r2r(mm6, mm5);/* src - dst -> mm5 */ |
2021 | pmulhw_r2r(mm0, mm5); /* mm5 * alpha -> mm5 */ |
2022 | /* 11 + 11 - 16 = 6 bits, so all the lower uninteresting |
2023 | bits are gone and the sign bits present */ |
2024 | psllw_i2r(5, mm5); /* mm5 << 5 -> mm5 */ |
2025 | paddw_r2r(mm5, mm6); /* mm5 + mm6(dst) -> mm6 */ |
2026 | |
2027 | por_r2r(mm6, mm1); /* save new greens in dsts */ |
2028 | |
2029 | /* blue */ |
2030 | movq_r2r(mm2, mm5); /* src -> mm5 */ |
2031 | movq_r2r(mm3, mm6); /* dst -> mm6 */ |
2032 | pand_r2r(mm7, mm5); /* src & MASKBLUE -> mm5[000b 000b 000b 000b] */ |
2033 | pand_r2r(mm7, mm6); /* dst & MASKBLUE -> mm6[000b 000b 000b 000b] */ |
2034 | |
2035 | /* blend */ |
2036 | psubw_r2r(mm6, mm5);/* src - dst -> mm5 */ |
2037 | pmullw_r2r(mm0, mm5); /* mm5 * alpha -> mm5 */ |
2038 | /* 11 + 5 = 16 bits, so the sign bits are lost and |
2039 | the interesting bits will need to be MASKed */ |
2040 | psrlw_i2r(11, mm5); /* mm5 >> 11 -> mm5 */ |
2041 | paddw_r2r(mm5, mm6); /* mm5 + mm6(dst) -> mm6 */ |
2042 | pand_r2r(mm7, mm6); /* mm6 & MASKBLUE -> mm6[000b 000b 000b 000b] */ |
2043 | |
2044 | por_r2r(mm6, mm1); /* save new blues in dsts */ |
2045 | |
2046 | movq_r2m(mm1, *dstp); /* mm1 -> 4 dst pixels */ |
2047 | |
2048 | srcp += 4; |
2049 | dstp += 4; |
2050 | }, width); |
2051 | srcp += srcskip; |
2052 | dstp += dstskip; |
2053 | } |
2054 | emms(); |
2055 | } |
2056 | } |
2057 | |
2058 | /* fast RGB555->RGB555 blending with surface alpha */ |
2059 | static void Blit555to555SurfaceAlphaMMX(SDL_BlitInfo *info) |
2060 | { |
2061 | unsigned alpha = info->src->alpha; /* downscale alpha to 5 bits */ |
2062 | if(alpha == 128) { |
2063 | Blit16to16SurfaceAlpha128(info, 0xfbde); |
2064 | } else { |
2065 | int width = info->d_width; |
2066 | int height = info->d_height; |
2067 | Uint16 *srcp = (Uint16 *)info->s_pixels; |
2068 | int srcskip = info->s_skip >> 1; |
2069 | Uint16 *dstp = (Uint16 *)info->d_pixels; |
2070 | int dstskip = info->d_skip >> 1; |
2071 | Uint32 s, d; |
2072 | Uint64 load; |
2073 | |
2074 | alpha &= ~(1+2+4); /* cut alpha to get the exact same behaviour */ |
2075 | load = alpha; |
2076 | alpha >>= 3; /* downscale alpha to 5 bits */ |
2077 | |
2078 | movq_m2r(load, mm0); /* alpha(0000000A) -> mm0 */ |
2079 | punpcklwd_r2r(mm0, mm0); /* 00000A0A -> mm0 */ |
2080 | punpcklwd_r2r(mm0, mm0); /* 0A0A0A0A -> mm0 */ |
2081 | /* position alpha to allow for mullo and mulhi on diff channels |
2082 | to reduce the number of operations */ |
2083 | psllq_i2r(3, mm0); |
2084 | |
2085 | /* Setup the 555 color channel masks */ |
2086 | load = 0x03E003E003E003E0ULL; |
2087 | movq_m2r(load, mm4); /* MASKGREEN -> mm4 */ |
2088 | load = 0x001F001F001F001FULL; |
2089 | movq_m2r(load, mm7); /* MASKBLUE -> mm7 */ |
2090 | while(height--) { |
2091 | DUFFS_LOOP_QUATRO2( |
2092 | { |
2093 | s = *srcp++; |
2094 | d = *dstp; |
2095 | /* |
2096 | * shift out the middle component (green) to |
2097 | * the high 16 bits, and process all three RGB |
2098 | * components at the same time. |
2099 | */ |
2100 | s = (s | s << 16) & 0x03e07c1f; |
2101 | d = (d | d << 16) & 0x03e07c1f; |
2102 | d += (s - d) * alpha >> 5; |
2103 | d &= 0x03e07c1f; |
2104 | *dstp++ = d | d >> 16; |
2105 | },{ |
2106 | s = *srcp++; |
2107 | d = *dstp; |
2108 | /* |
2109 | * shift out the middle component (green) to |
2110 | * the high 16 bits, and process all three RGB |
2111 | * components at the same time. |
2112 | */ |
2113 | s = (s | s << 16) & 0x03e07c1f; |
2114 | d = (d | d << 16) & 0x03e07c1f; |
2115 | d += (s - d) * alpha >> 5; |
2116 | d &= 0x03e07c1f; |
2117 | *dstp++ = d | d >> 16; |
2118 | s = *srcp++; |
2119 | d = *dstp; |
2120 | /* |
2121 | * shift out the middle component (green) to |
2122 | * the high 16 bits, and process all three RGB |
2123 | * components at the same time. |
2124 | */ |
2125 | s = (s | s << 16) & 0x03e07c1f; |
2126 | d = (d | d << 16) & 0x03e07c1f; |
2127 | d += (s - d) * alpha >> 5; |
2128 | d &= 0x03e07c1f; |
2129 | *dstp++ = d | d >> 16; |
2130 | },{ |
2131 | movq_m2r((*srcp), mm2);/* 4 src pixels -> mm2 */ |
2132 | movq_m2r((*dstp), mm3);/* 4 dst pixels -> mm3 */ |
2133 | |
2134 | /* red -- process the bits in place */ |
2135 | psllq_i2r(5, mm4); /* turn MASKGREEN into MASKRED */ |
2136 | /* by reusing the GREEN mask we free up another mmx |
2137 | register to accumulate the result */ |
2138 | |
2139 | movq_r2r(mm2, mm5); /* src -> mm5 */ |
2140 | movq_r2r(mm3, mm6); /* dst -> mm6 */ |
2141 | pand_r2r(mm4, mm5); /* src & MASKRED -> mm5 */ |
2142 | pand_r2r(mm4, mm6); /* dst & MASKRED -> mm6 */ |
2143 | |
2144 | /* blend */ |
2145 | psubw_r2r(mm6, mm5);/* src - dst -> mm5 */ |
2146 | pmulhw_r2r(mm0, mm5); /* mm5 * alpha -> mm5 */ |
2147 | /* 11 + 15 - 16 = 10 bits, uninteresting bits will be |
2148 | cleared by a MASK below */ |
2149 | psllw_i2r(5, mm5); /* mm5 << 5 -> mm5 */ |
2150 | paddw_r2r(mm5, mm6); /* mm5 + mm6(dst) -> mm6 */ |
2151 | pand_r2r(mm4, mm6); /* mm6 & MASKRED -> mm6 */ |
2152 | |
2153 | psrlq_i2r(5, mm4); /* turn MASKRED back into MASKGREEN */ |
2154 | |
2155 | movq_r2r(mm6, mm1); /* save new reds in dsts */ |
2156 | |
2157 | /* green -- process the bits in place */ |
2158 | movq_r2r(mm2, mm5); /* src -> mm5 */ |
2159 | movq_r2r(mm3, mm6); /* dst -> mm6 */ |
2160 | pand_r2r(mm4, mm5); /* src & MASKGREEN -> mm5 */ |
2161 | pand_r2r(mm4, mm6); /* dst & MASKGREEN -> mm6 */ |
2162 | |
2163 | /* blend */ |
2164 | psubw_r2r(mm6, mm5);/* src - dst -> mm5 */ |
2165 | pmulhw_r2r(mm0, mm5); /* mm5 * alpha -> mm5 */ |
2166 | /* 11 + 10 - 16 = 5 bits, so all the lower uninteresting |
2167 | bits are gone and the sign bits present */ |
2168 | psllw_i2r(5, mm5); /* mm5 << 5 -> mm5 */ |
2169 | paddw_r2r(mm5, mm6); /* mm5 + mm6(dst) -> mm6 */ |
2170 | |
2171 | por_r2r(mm6, mm1); /* save new greens in dsts */ |
2172 | |
2173 | /* blue */ |
2174 | movq_r2r(mm2, mm5); /* src -> mm5 */ |
2175 | movq_r2r(mm3, mm6); /* dst -> mm6 */ |
2176 | pand_r2r(mm7, mm5); /* src & MASKBLUE -> mm5[000b 000b 000b 000b] */ |
2177 | pand_r2r(mm7, mm6); /* dst & MASKBLUE -> mm6[000b 000b 000b 000b] */ |
2178 | |
2179 | /* blend */ |
2180 | psubw_r2r(mm6, mm5);/* src - dst -> mm5 */ |
2181 | pmullw_r2r(mm0, mm5); /* mm5 * alpha -> mm5 */ |
2182 | /* 11 + 5 = 16 bits, so the sign bits are lost and |
2183 | the interesting bits will need to be MASKed */ |
2184 | psrlw_i2r(11, mm5); /* mm5 >> 11 -> mm5 */ |
2185 | paddw_r2r(mm5, mm6); /* mm5 + mm6(dst) -> mm6 */ |
2186 | pand_r2r(mm7, mm6); /* mm6 & MASKBLUE -> mm6[000b 000b 000b 000b] */ |
2187 | |
2188 | por_r2r(mm6, mm1); /* save new blues in dsts */ |
2189 | |
2190 | movq_r2m(mm1, *dstp);/* mm1 -> 4 dst pixels */ |
2191 | |
2192 | srcp += 4; |
2193 | dstp += 4; |
2194 | }, width); |
2195 | srcp += srcskip; |
2196 | dstp += dstskip; |
2197 | } |
2198 | emms(); |
2199 | } |
2200 | } |
2201 | /* End GCC_ASMBLIT */ |
2202 | |
2203 | #elif MSVC_ASMBLIT |
2204 | /* fast RGB565->RGB565 blending with surface alpha */ |
2205 | static void Blit565to565SurfaceAlphaMMX(SDL_BlitInfo *info) |
2206 | { |
2207 | unsigned alpha = info->src->alpha; |
2208 | if(alpha == 128) { |
2209 | Blit16to16SurfaceAlpha128(info, 0xf7de); |
2210 | } else { |
2211 | int width = info->d_width; |
2212 | int height = info->d_height; |
2213 | Uint16 *srcp = (Uint16 *)info->s_pixels; |
2214 | int srcskip = info->s_skip >> 1; |
2215 | Uint16 *dstp = (Uint16 *)info->d_pixels; |
2216 | int dstskip = info->d_skip >> 1; |
2217 | Uint32 s, d; |
2218 | |
2219 | __m64 src1, dst1, src2, dst2, gmask, bmask, mm_res, mm_alpha; |
2220 | |
2221 | alpha &= ~(1+2+4); /* cut alpha to get the exact same behaviour */ |
2222 | mm_alpha = _mm_set_pi32(0, alpha); /* 0000000A -> mm_alpha */ |
2223 | alpha >>= 3; /* downscale alpha to 5 bits */ |
2224 | |
2225 | mm_alpha = _mm_unpacklo_pi16(mm_alpha, mm_alpha); /* 00000A0A -> mm_alpha */ |
2226 | mm_alpha = _mm_unpacklo_pi32(mm_alpha, mm_alpha); /* 0A0A0A0A -> mm_alpha */ |
2227 | /* position alpha to allow for mullo and mulhi on diff channels |
2228 | to reduce the number of operations */ |
2229 | mm_alpha = _mm_slli_si64(mm_alpha, 3); |
2230 | |
2231 | /* Setup the 565 color channel masks */ |
2232 | gmask = _mm_set_pi32(0x07E007E0, 0x07E007E0); /* MASKGREEN -> gmask */ |
2233 | bmask = _mm_set_pi32(0x001F001F, 0x001F001F); /* MASKBLUE -> bmask */ |
2234 | |
2235 | while(height--) { |
2236 | DUFFS_LOOP_QUATRO2( |
2237 | { |
2238 | s = *srcp++; |
2239 | d = *dstp; |
2240 | /* |
2241 | * shift out the middle component (green) to |
2242 | * the high 16 bits, and process all three RGB |
2243 | * components at the same time. |
2244 | */ |
2245 | s = (s | s << 16) & 0x07e0f81f; |
2246 | d = (d | d << 16) & 0x07e0f81f; |
2247 | d += (s - d) * alpha >> 5; |
2248 | d &= 0x07e0f81f; |
2249 | *dstp++ = (Uint16)(d | d >> 16); |
2250 | },{ |
2251 | s = *srcp++; |
2252 | d = *dstp; |
2253 | /* |
2254 | * shift out the middle component (green) to |
2255 | * the high 16 bits, and process all three RGB |
2256 | * components at the same time. |
2257 | */ |
2258 | s = (s | s << 16) & 0x07e0f81f; |
2259 | d = (d | d << 16) & 0x07e0f81f; |
2260 | d += (s - d) * alpha >> 5; |
2261 | d &= 0x07e0f81f; |
2262 | *dstp++ = (Uint16)(d | d >> 16); |
2263 | s = *srcp++; |
2264 | d = *dstp; |
2265 | /* |
2266 | * shift out the middle component (green) to |
2267 | * the high 16 bits, and process all three RGB |
2268 | * components at the same time. |
2269 | */ |
2270 | s = (s | s << 16) & 0x07e0f81f; |
2271 | d = (d | d << 16) & 0x07e0f81f; |
2272 | d += (s - d) * alpha >> 5; |
2273 | d &= 0x07e0f81f; |
2274 | *dstp++ = (Uint16)(d | d >> 16); |
2275 | },{ |
2276 | src1 = *(__m64*)srcp; /* 4 src pixels -> src1 */ |
2277 | dst1 = *(__m64*)dstp; /* 4 dst pixels -> dst1 */ |
2278 | |
2279 | /* red */ |
2280 | src2 = src1; |
2281 | src2 = _mm_srli_pi16(src2, 11); /* src2 >> 11 -> src2 [000r 000r 000r 000r] */ |
2282 | |
2283 | dst2 = dst1; |
2284 | dst2 = _mm_srli_pi16(dst2, 11); /* dst2 >> 11 -> dst2 [000r 000r 000r 000r] */ |
2285 | |
2286 | /* blend */ |
2287 | src2 = _mm_sub_pi16(src2, dst2);/* src - dst -> src2 */ |
2288 | src2 = _mm_mullo_pi16(src2, mm_alpha); /* src2 * alpha -> src2 */ |
2289 | src2 = _mm_srli_pi16(src2, 11); /* src2 >> 11 -> src2 */ |
2290 | dst2 = _mm_add_pi16(src2, dst2); /* src2 + dst2 -> dst2 */ |
2291 | dst2 = _mm_slli_pi16(dst2, 11); /* dst2 << 11 -> dst2 */ |
2292 | |
2293 | mm_res = dst2; /* RED -> mm_res */ |
2294 | |
2295 | /* green -- process the bits in place */ |
2296 | src2 = src1; |
2297 | src2 = _mm_and_si64(src2, gmask); /* src & MASKGREEN -> src2 */ |
2298 | |
2299 | dst2 = dst1; |
2300 | dst2 = _mm_and_si64(dst2, gmask); /* dst & MASKGREEN -> dst2 */ |
2301 | |
2302 | /* blend */ |
2303 | src2 = _mm_sub_pi16(src2, dst2);/* src - dst -> src2 */ |
2304 | src2 = _mm_mulhi_pi16(src2, mm_alpha); /* src2 * alpha -> src2 */ |
2305 | src2 = _mm_slli_pi16(src2, 5); /* src2 << 5 -> src2 */ |
2306 | dst2 = _mm_add_pi16(src2, dst2); /* src2 + dst2 -> dst2 */ |
2307 | |
2308 | mm_res = _mm_or_si64(mm_res, dst2); /* RED | GREEN -> mm_res */ |
2309 | |
2310 | /* blue */ |
2311 | src2 = src1; |
2312 | src2 = _mm_and_si64(src2, bmask); /* src & MASKBLUE -> src2[000b 000b 000b 000b] */ |
2313 | |
2314 | dst2 = dst1; |
2315 | dst2 = _mm_and_si64(dst2, bmask); /* dst & MASKBLUE -> dst2[000b 000b 000b 000b] */ |
2316 | |
2317 | /* blend */ |
2318 | src2 = _mm_sub_pi16(src2, dst2);/* src - dst -> src2 */ |
2319 | src2 = _mm_mullo_pi16(src2, mm_alpha); /* src2 * alpha -> src2 */ |
2320 | src2 = _mm_srli_pi16(src2, 11); /* src2 >> 11 -> src2 */ |
2321 | dst2 = _mm_add_pi16(src2, dst2); /* src2 + dst2 -> dst2 */ |
2322 | dst2 = _mm_and_si64(dst2, bmask); /* dst2 & MASKBLUE -> dst2 */ |
2323 | |
2324 | mm_res = _mm_or_si64(mm_res, dst2); /* RED | GREEN | BLUE -> mm_res */ |
2325 | |
2326 | *(__m64*)dstp = mm_res; /* mm_res -> 4 dst pixels */ |
2327 | |
2328 | srcp += 4; |
2329 | dstp += 4; |
2330 | }, width); |
2331 | srcp += srcskip; |
2332 | dstp += dstskip; |
2333 | } |
2334 | _mm_empty(); |
2335 | } |
2336 | } |
2337 | |
2338 | /* fast RGB555->RGB555 blending with surface alpha */ |
2339 | static void Blit555to555SurfaceAlphaMMX(SDL_BlitInfo *info) |
2340 | { |
2341 | unsigned alpha = info->src->alpha; |
2342 | if(alpha == 128) { |
2343 | Blit16to16SurfaceAlpha128(info, 0xfbde); |
2344 | } else { |
2345 | int width = info->d_width; |
2346 | int height = info->d_height; |
2347 | Uint16 *srcp = (Uint16 *)info->s_pixels; |
2348 | int srcskip = info->s_skip >> 1; |
2349 | Uint16 *dstp = (Uint16 *)info->d_pixels; |
2350 | int dstskip = info->d_skip >> 1; |
2351 | Uint32 s, d; |
2352 | |
2353 | __m64 src1, dst1, src2, dst2, rmask, gmask, bmask, mm_res, mm_alpha; |
2354 | |
2355 | alpha &= ~(1+2+4); /* cut alpha to get the exact same behaviour */ |
2356 | mm_alpha = _mm_set_pi32(0, alpha); /* 0000000A -> mm_alpha */ |
2357 | alpha >>= 3; /* downscale alpha to 5 bits */ |
2358 | |
2359 | mm_alpha = _mm_unpacklo_pi16(mm_alpha, mm_alpha); /* 00000A0A -> mm_alpha */ |
2360 | mm_alpha = _mm_unpacklo_pi32(mm_alpha, mm_alpha); /* 0A0A0A0A -> mm_alpha */ |
2361 | /* position alpha to allow for mullo and mulhi on diff channels |
2362 | to reduce the number of operations */ |
2363 | mm_alpha = _mm_slli_si64(mm_alpha, 3); |
2364 | |
2365 | /* Setup the 555 color channel masks */ |
2366 | rmask = _mm_set_pi32(0x7C007C00, 0x7C007C00); /* MASKRED -> rmask */ |
2367 | gmask = _mm_set_pi32(0x03E003E0, 0x03E003E0); /* MASKGREEN -> gmask */ |
2368 | bmask = _mm_set_pi32(0x001F001F, 0x001F001F); /* MASKBLUE -> bmask */ |
2369 | |
2370 | while(height--) { |
2371 | DUFFS_LOOP_QUATRO2( |
2372 | { |
2373 | s = *srcp++; |
2374 | d = *dstp; |
2375 | /* |
2376 | * shift out the middle component (green) to |
2377 | * the high 16 bits, and process all three RGB |
2378 | * components at the same time. |
2379 | */ |
2380 | s = (s | s << 16) & 0x03e07c1f; |
2381 | d = (d | d << 16) & 0x03e07c1f; |
2382 | d += (s - d) * alpha >> 5; |
2383 | d &= 0x03e07c1f; |
2384 | *dstp++ = (Uint16)(d | d >> 16); |
2385 | },{ |
2386 | s = *srcp++; |
2387 | d = *dstp; |
2388 | /* |
2389 | * shift out the middle component (green) to |
2390 | * the high 16 bits, and process all three RGB |
2391 | * components at the same time. |
2392 | */ |
2393 | s = (s | s << 16) & 0x03e07c1f; |
2394 | d = (d | d << 16) & 0x03e07c1f; |
2395 | d += (s - d) * alpha >> 5; |
2396 | d &= 0x03e07c1f; |
2397 | *dstp++ = (Uint16)(d | d >> 16); |
2398 | s = *srcp++; |
2399 | d = *dstp; |
2400 | /* |
2401 | * shift out the middle component (green) to |
2402 | * the high 16 bits, and process all three RGB |
2403 | * components at the same time. |
2404 | */ |
2405 | s = (s | s << 16) & 0x03e07c1f; |
2406 | d = (d | d << 16) & 0x03e07c1f; |
2407 | d += (s - d) * alpha >> 5; |
2408 | d &= 0x03e07c1f; |
2409 | *dstp++ = (Uint16)(d | d >> 16); |
2410 | },{ |
2411 | src1 = *(__m64*)srcp; /* 4 src pixels -> src1 */ |
2412 | dst1 = *(__m64*)dstp; /* 4 dst pixels -> dst1 */ |
2413 | |
2414 | /* red -- process the bits in place */ |
2415 | src2 = src1; |
2416 | src2 = _mm_and_si64(src2, rmask); /* src & MASKRED -> src2 */ |
2417 | |
2418 | dst2 = dst1; |
2419 | dst2 = _mm_and_si64(dst2, rmask); /* dst & MASKRED -> dst2 */ |
2420 | |
2421 | /* blend */ |
2422 | src2 = _mm_sub_pi16(src2, dst2);/* src - dst -> src2 */ |
2423 | src2 = _mm_mulhi_pi16(src2, mm_alpha); /* src2 * alpha -> src2 */ |
2424 | src2 = _mm_slli_pi16(src2, 5); /* src2 << 5 -> src2 */ |
2425 | dst2 = _mm_add_pi16(src2, dst2); /* src2 + dst2 -> dst2 */ |
2426 | dst2 = _mm_and_si64(dst2, rmask); /* dst2 & MASKRED -> dst2 */ |
2427 | |
2428 | mm_res = dst2; /* RED -> mm_res */ |
2429 | |
2430 | /* green -- process the bits in place */ |
2431 | src2 = src1; |
2432 | src2 = _mm_and_si64(src2, gmask); /* src & MASKGREEN -> src2 */ |
2433 | |
2434 | dst2 = dst1; |
2435 | dst2 = _mm_and_si64(dst2, gmask); /* dst & MASKGREEN -> dst2 */ |
2436 | |
2437 | /* blend */ |
2438 | src2 = _mm_sub_pi16(src2, dst2);/* src - dst -> src2 */ |
2439 | src2 = _mm_mulhi_pi16(src2, mm_alpha); /* src2 * alpha -> src2 */ |
2440 | src2 = _mm_slli_pi16(src2, 5); /* src2 << 5 -> src2 */ |
2441 | dst2 = _mm_add_pi16(src2, dst2); /* src2 + dst2 -> dst2 */ |
2442 | |
2443 | mm_res = _mm_or_si64(mm_res, dst2); /* RED | GREEN -> mm_res */ |
2444 | |
2445 | /* blue */ |
2446 | src2 = src1; /* src -> src2 */ |
2447 | src2 = _mm_and_si64(src2, bmask); /* src & MASKBLUE -> src2[000b 000b 000b 000b] */ |
2448 | |
2449 | dst2 = dst1; /* dst -> dst2 */ |
2450 | dst2 = _mm_and_si64(dst2, bmask); /* dst & MASKBLUE -> dst2[000b 000b 000b 000b] */ |
2451 | |
2452 | /* blend */ |
2453 | src2 = _mm_sub_pi16(src2, dst2);/* src - dst -> src2 */ |
2454 | src2 = _mm_mullo_pi16(src2, mm_alpha); /* src2 * alpha -> src2 */ |
2455 | src2 = _mm_srli_pi16(src2, 11); /* src2 >> 11 -> src2 */ |
2456 | dst2 = _mm_add_pi16(src2, dst2); /* src2 + dst2 -> dst2 */ |
2457 | dst2 = _mm_and_si64(dst2, bmask); /* dst2 & MASKBLUE -> dst2 */ |
2458 | |
2459 | mm_res = _mm_or_si64(mm_res, dst2); /* RED | GREEN | BLUE -> mm_res */ |
2460 | |
2461 | *(__m64*)dstp = mm_res; /* mm_res -> 4 dst pixels */ |
2462 | |
2463 | srcp += 4; |
2464 | dstp += 4; |
2465 | }, width); |
2466 | srcp += srcskip; |
2467 | dstp += dstskip; |
2468 | } |
2469 | _mm_empty(); |
2470 | } |
2471 | } |
2472 | #endif /* GCC_ASMBLIT, MSVC_ASMBLIT */ |
2473 | |
2474 | /* fast RGB565->RGB565 blending with surface alpha */ |
2475 | static void Blit565to565SurfaceAlpha(SDL_BlitInfo *info) |
2476 | { |
2477 | unsigned alpha = info->src->alpha; |
2478 | if(alpha == 128) { |
2479 | Blit16to16SurfaceAlpha128(info, 0xf7de); |
2480 | } else { |
2481 | int width = info->d_width; |
2482 | int height = info->d_height; |
2483 | Uint16 *srcp = (Uint16 *)info->s_pixels; |
2484 | int srcskip = info->s_skip >> 1; |
2485 | Uint16 *dstp = (Uint16 *)info->d_pixels; |
2486 | int dstskip = info->d_skip >> 1; |
2487 | alpha >>= 3; /* downscale alpha to 5 bits */ |
2488 | |
2489 | while(height--) { |
2490 | DUFFS_LOOP4({ |
2491 | Uint32 s = *srcp++; |
2492 | Uint32 d = *dstp; |
2493 | /* |
2494 | * shift out the middle component (green) to |
2495 | * the high 16 bits, and process all three RGB |
2496 | * components at the same time. |
2497 | */ |
2498 | s = (s | s << 16) & 0x07e0f81f; |
2499 | d = (d | d << 16) & 0x07e0f81f; |
2500 | d += (s - d) * alpha >> 5; |
2501 | d &= 0x07e0f81f; |
2502 | *dstp++ = (Uint16)(d | d >> 16); |
2503 | }, width); |
2504 | srcp += srcskip; |
2505 | dstp += dstskip; |
2506 | } |
2507 | } |
2508 | } |
2509 | |
2510 | /* fast RGB555->RGB555 blending with surface alpha */ |
2511 | static void Blit555to555SurfaceAlpha(SDL_BlitInfo *info) |
2512 | { |
2513 | unsigned alpha = info->src->alpha; /* downscale alpha to 5 bits */ |
2514 | if(alpha == 128) { |
2515 | Blit16to16SurfaceAlpha128(info, 0xfbde); |
2516 | } else { |
2517 | int width = info->d_width; |
2518 | int height = info->d_height; |
2519 | Uint16 *srcp = (Uint16 *)info->s_pixels; |
2520 | int srcskip = info->s_skip >> 1; |
2521 | Uint16 *dstp = (Uint16 *)info->d_pixels; |
2522 | int dstskip = info->d_skip >> 1; |
2523 | alpha >>= 3; /* downscale alpha to 5 bits */ |
2524 | |
2525 | while(height--) { |
2526 | DUFFS_LOOP4({ |
2527 | Uint32 s = *srcp++; |
2528 | Uint32 d = *dstp; |
2529 | /* |
2530 | * shift out the middle component (green) to |
2531 | * the high 16 bits, and process all three RGB |
2532 | * components at the same time. |
2533 | */ |
2534 | s = (s | s << 16) & 0x03e07c1f; |
2535 | d = (d | d << 16) & 0x03e07c1f; |
2536 | d += (s - d) * alpha >> 5; |
2537 | d &= 0x03e07c1f; |
2538 | *dstp++ = (Uint16)(d | d >> 16); |
2539 | }, width); |
2540 | srcp += srcskip; |
2541 | dstp += dstskip; |
2542 | } |
2543 | } |
2544 | } |
2545 | |
2546 | /* fast ARGB8888->RGB565 blending with pixel alpha */ |
2547 | static void BlitARGBto565PixelAlpha(SDL_BlitInfo *info) |
2548 | { |
2549 | int width = info->d_width; |
2550 | int height = info->d_height; |
2551 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
2552 | int srcskip = info->s_skip >> 2; |
2553 | Uint16 *dstp = (Uint16 *)info->d_pixels; |
2554 | int dstskip = info->d_skip >> 1; |
2555 | |
2556 | while(height--) { |
2557 | DUFFS_LOOP4({ |
2558 | Uint32 s = *srcp; |
2559 | unsigned alpha = s >> 27; /* downscale alpha to 5 bits */ |
2560 | /* FIXME: Here we special-case opaque alpha since the |
2561 | compositioning used (>>8 instead of /255) doesn't handle |
2562 | it correctly. Also special-case alpha=0 for speed? |
2563 | Benchmark this! */ |
2564 | if(alpha) { |
2565 | if(alpha == (SDL_ALPHA_OPAQUE >> 3)) { |
2566 | *dstp = (Uint16)((s >> 8 & 0xf800) + (s >> 5 & 0x7e0) + (s >> 3 & 0x1f)); |
2567 | } else { |
2568 | Uint32 d = *dstp; |
2569 | /* |
2570 | * convert source and destination to G0RAB65565 |
2571 | * and blend all components at the same time |
2572 | */ |
2573 | s = ((s & 0xfc00) << 11) + (s >> 8 & 0xf800) |
2574 | + (s >> 3 & 0x1f); |
2575 | d = (d | d << 16) & 0x07e0f81f; |
2576 | d += (s - d) * alpha >> 5; |
2577 | d &= 0x07e0f81f; |
2578 | *dstp = (Uint16)(d | d >> 16); |
2579 | } |
2580 | } |
2581 | srcp++; |
2582 | dstp++; |
2583 | }, width); |
2584 | srcp += srcskip; |
2585 | dstp += dstskip; |
2586 | } |
2587 | } |
2588 | |
2589 | /* fast ARGB8888->RGB555 blending with pixel alpha */ |
2590 | static void BlitARGBto555PixelAlpha(SDL_BlitInfo *info) |
2591 | { |
2592 | int width = info->d_width; |
2593 | int height = info->d_height; |
2594 | Uint32 *srcp = (Uint32 *)info->s_pixels; |
2595 | int srcskip = info->s_skip >> 2; |
2596 | Uint16 *dstp = (Uint16 *)info->d_pixels; |
2597 | int dstskip = info->d_skip >> 1; |
2598 | |
2599 | while(height--) { |
2600 | DUFFS_LOOP4({ |
2601 | unsigned alpha; |
2602 | Uint32 s = *srcp; |
2603 | alpha = s >> 27; /* downscale alpha to 5 bits */ |
2604 | /* FIXME: Here we special-case opaque alpha since the |
2605 | compositioning used (>>8 instead of /255) doesn't handle |
2606 | it correctly. Also special-case alpha=0 for speed? |
2607 | Benchmark this! */ |
2608 | if(alpha) { |
2609 | if(alpha == (SDL_ALPHA_OPAQUE >> 3)) { |
2610 | *dstp = (Uint16)((s >> 9 & 0x7c00) + (s >> 6 & 0x3e0) + (s >> 3 & 0x1f)); |
2611 | } else { |
2612 | Uint32 d = *dstp; |
2613 | /* |
2614 | * convert source and destination to G0RAB65565 |
2615 | * and blend all components at the same time |
2616 | */ |
2617 | s = ((s & 0xf800) << 10) + (s >> 9 & 0x7c00) |
2618 | + (s >> 3 & 0x1f); |
2619 | d = (d | d << 16) & 0x03e07c1f; |
2620 | d += (s - d) * alpha >> 5; |
2621 | d &= 0x03e07c1f; |
2622 | *dstp = (Uint16)(d | d >> 16); |
2623 | } |
2624 | } |
2625 | srcp++; |
2626 | dstp++; |
2627 | }, width); |
2628 | srcp += srcskip; |
2629 | dstp += dstskip; |
2630 | } |
2631 | } |
2632 | |
2633 | /* General (slow) N->N blending with per-surface alpha */ |
2634 | static void BlitNtoNSurfaceAlpha(SDL_BlitInfo *info) |
2635 | { |
2636 | int width = info->d_width; |
2637 | int height = info->d_height; |
2638 | Uint8 *src = info->s_pixels; |
2639 | int srcskip = info->s_skip; |
2640 | Uint8 *dst = info->d_pixels; |
2641 | int dstskip = info->d_skip; |
2642 | SDL_PixelFormat *srcfmt = info->src; |
2643 | SDL_PixelFormat *dstfmt = info->dst; |
2644 | int srcbpp = srcfmt->BytesPerPixel; |
2645 | int dstbpp = dstfmt->BytesPerPixel; |
2646 | unsigned sA = srcfmt->alpha; |
2647 | unsigned dA = dstfmt->Amask ? SDL_ALPHA_OPAQUE : 0; |
2648 | |
2649 | if(sA) { |
2650 | while ( height-- ) { |
2651 | DUFFS_LOOP4( |
2652 | { |
2653 | Uint32 Pixel; |
2654 | unsigned sR; |
2655 | unsigned sG; |
2656 | unsigned sB; |
2657 | unsigned dR; |
2658 | unsigned dG; |
2659 | unsigned dB; |
2660 | DISEMBLE_RGB(src, srcbpp, srcfmt, Pixel, sR, sG, sB); |
2661 | DISEMBLE_RGB(dst, dstbpp, dstfmt, Pixel, dR, dG, dB); |
2662 | ALPHA_BLEND(sR, sG, sB, sA, dR, dG, dB); |
2663 | ASSEMBLE_RGBA(dst, dstbpp, dstfmt, dR, dG, dB, dA); |
2664 | src += srcbpp; |
2665 | dst += dstbpp; |
2666 | }, |
2667 | width); |
2668 | src += srcskip; |
2669 | dst += dstskip; |
2670 | } |
2671 | } |
2672 | } |
2673 | |
2674 | /* General (slow) colorkeyed N->N blending with per-surface alpha */ |
2675 | static void BlitNtoNSurfaceAlphaKey(SDL_BlitInfo *info) |
2676 | { |
2677 | int width = info->d_width; |
2678 | int height = info->d_height; |
2679 | Uint8 *src = info->s_pixels; |
2680 | int srcskip = info->s_skip; |
2681 | Uint8 *dst = info->d_pixels; |
2682 | int dstskip = info->d_skip; |
2683 | SDL_PixelFormat *srcfmt = info->src; |
2684 | SDL_PixelFormat *dstfmt = info->dst; |
2685 | Uint32 ckey = srcfmt->colorkey; |
2686 | int srcbpp = srcfmt->BytesPerPixel; |
2687 | int dstbpp = dstfmt->BytesPerPixel; |
2688 | unsigned sA = srcfmt->alpha; |
2689 | unsigned dA = dstfmt->Amask ? SDL_ALPHA_OPAQUE : 0; |
2690 | |
211e4bff |
2691 | if (srcbpp == 2 && srcfmt->Gmask == 0x7e0 && dstbpp == 2 && dstfmt->Gmask == 0x7e0) { |
2692 | Uint16 *src16 = (Uint16 *)src; |
2693 | Uint16 *dst16 = (Uint16 *)dst; |
2694 | sA >>= 3; /* downscale alpha to 5 bits */ |
2695 | while ( height-- ) { |
2696 | DUFFS_LOOP4( |
2697 | { |
2698 | Uint32 s; |
2699 | Uint32 d; |
2700 | s = *src16; |
2701 | if(sA && s != ckey) { |
2702 | d = *dst16; |
2703 | s = (s | s << 16) & 0x07e0f81f; |
2704 | d = (d | d << 16) & 0x07e0f81f; |
2705 | d += (s - d) * sA >> 5; |
2706 | d &= 0x07e0f81f; |
2707 | *dst16 = (Uint16)(d | d >> 16); |
2708 | } |
2709 | src16++; |
2710 | dst16++; |
2711 | }, |
2712 | width); |
2713 | src16 += srcskip / 2; |
2714 | dst16 += dstskip / 2; |
2715 | } |
2716 | return; |
2717 | } |
2718 | |
e14743d1 |
2719 | while ( height-- ) { |
2720 | DUFFS_LOOP4( |
2721 | { |
2722 | Uint32 Pixel; |
2723 | unsigned sR; |
2724 | unsigned sG; |
2725 | unsigned sB; |
2726 | unsigned dR; |
2727 | unsigned dG; |
2728 | unsigned dB; |
2729 | RETRIEVE_RGB_PIXEL(src, srcbpp, Pixel); |
2730 | if(sA && Pixel != ckey) { |
2731 | RGB_FROM_PIXEL(Pixel, srcfmt, sR, sG, sB); |
2732 | DISEMBLE_RGB(dst, dstbpp, dstfmt, Pixel, dR, dG, dB); |
2733 | ALPHA_BLEND(sR, sG, sB, sA, dR, dG, dB); |
2734 | ASSEMBLE_RGBA(dst, dstbpp, dstfmt, dR, dG, dB, dA); |
2735 | } |
2736 | src += srcbpp; |
2737 | dst += dstbpp; |
2738 | }, |
2739 | width); |
2740 | src += srcskip; |
2741 | dst += dstskip; |
2742 | } |
2743 | } |
2744 | |
2745 | /* General (slow) N->N blending with pixel alpha */ |
2746 | static void BlitNtoNPixelAlpha(SDL_BlitInfo *info) |
2747 | { |
2748 | int width = info->d_width; |
2749 | int height = info->d_height; |
2750 | Uint8 *src = info->s_pixels; |
2751 | int srcskip = info->s_skip; |
2752 | Uint8 *dst = info->d_pixels; |
2753 | int dstskip = info->d_skip; |
2754 | SDL_PixelFormat *srcfmt = info->src; |
2755 | SDL_PixelFormat *dstfmt = info->dst; |
2756 | |
2757 | int srcbpp; |
2758 | int dstbpp; |
2759 | |
2760 | /* Set up some basic variables */ |
2761 | srcbpp = srcfmt->BytesPerPixel; |
2762 | dstbpp = dstfmt->BytesPerPixel; |
2763 | |
2764 | /* FIXME: for 8bpp source alpha, this doesn't get opaque values |
2765 | quite right. for <8bpp source alpha, it gets them very wrong |
2766 | (check all macros!) |
2767 | It is unclear whether there is a good general solution that doesn't |
2768 | need a branch (or a divide). */ |
2769 | while ( height-- ) { |
2770 | DUFFS_LOOP4( |
2771 | { |
2772 | Uint32 Pixel; |
2773 | unsigned sR; |
2774 | unsigned sG; |
2775 | unsigned sB; |
2776 | unsigned dR; |
2777 | unsigned dG; |
2778 | unsigned dB; |
2779 | unsigned sA; |
2780 | unsigned dA; |
2781 | DISEMBLE_RGBA(src, srcbpp, srcfmt, Pixel, sR, sG, sB, sA); |
2782 | if(sA) { |
2783 | DISEMBLE_RGBA(dst, dstbpp, dstfmt, Pixel, dR, dG, dB, dA); |
2784 | ALPHA_BLEND(sR, sG, sB, sA, dR, dG, dB); |
2785 | ASSEMBLE_RGBA(dst, dstbpp, dstfmt, dR, dG, dB, dA); |
2786 | } |
2787 | src += srcbpp; |
2788 | dst += dstbpp; |
2789 | }, |
2790 | width); |
2791 | src += srcskip; |
2792 | dst += dstskip; |
2793 | } |
2794 | } |
2795 | |
2796 | |
2797 | SDL_loblit SDL_CalculateAlphaBlit(SDL_Surface *surface, int blit_index) |
2798 | { |
2799 | SDL_PixelFormat *sf = surface->format; |
2800 | SDL_PixelFormat *df = surface->map->dst->format; |
2801 | |
2802 | if(sf->Amask == 0) { |
2803 | if((surface->flags & SDL_SRCCOLORKEY) == SDL_SRCCOLORKEY) { |
2804 | if(df->BytesPerPixel == 1) |
2805 | return BlitNto1SurfaceAlphaKey; |
2806 | else |
2807 | #if SDL_ALTIVEC_BLITTERS |
2808 | if (sf->BytesPerPixel == 4 && df->BytesPerPixel == 4 && |
2809 | !(surface->map->dst->flags & SDL_HWSURFACE) && SDL_HasAltiVec()) |
2810 | return Blit32to32SurfaceAlphaKeyAltivec; |
2811 | else |
2812 | #endif |
2813 | return BlitNtoNSurfaceAlphaKey; |
2814 | } else { |
2815 | /* Per-surface alpha blits */ |
2816 | switch(df->BytesPerPixel) { |
2817 | case 1: |
2818 | return BlitNto1SurfaceAlpha; |
2819 | |
2820 | case 2: |
2821 | if(surface->map->identity) { |
2822 | if(df->Gmask == 0x7e0) |
2823 | { |
2824 | #if MMX_ASMBLIT |
2825 | if(SDL_HasMMX()) |
2826 | return Blit565to565SurfaceAlphaMMX; |
2827 | else |
2828 | #endif |
2829 | return Blit565to565SurfaceAlpha; |
2830 | } |
2831 | else if(df->Gmask == 0x3e0) |
2832 | { |
2833 | #if MMX_ASMBLIT |
2834 | if(SDL_HasMMX()) |
2835 | return Blit555to555SurfaceAlphaMMX; |
2836 | else |
2837 | #endif |
2838 | return Blit555to555SurfaceAlpha; |
2839 | } |
2840 | } |
2841 | return BlitNtoNSurfaceAlpha; |
2842 | |
2843 | case 4: |
2844 | if(sf->Rmask == df->Rmask |
2845 | && sf->Gmask == df->Gmask |
2846 | && sf->Bmask == df->Bmask |
2847 | && sf->BytesPerPixel == 4) |
2848 | { |
2849 | #if MMX_ASMBLIT |
2850 | if(sf->Rshift % 8 == 0 |
2851 | && sf->Gshift % 8 == 0 |
2852 | && sf->Bshift % 8 == 0 |
2853 | && SDL_HasMMX()) |
2854 | return BlitRGBtoRGBSurfaceAlphaMMX; |
bdfa6989 |
2855 | #endif |
2856 | #ifdef __ARM_NEON__ |
2857 | if(sf->Rshift % 8 == 0 |
2858 | && sf->Gshift % 8 == 0 |
2859 | && sf->Bshift % 8 == 0) |
c85a5291 |
2860 | { |
bdfa6989 |
2861 | return BlitARGBtoXRGBalphaS_neon; |
c85a5291 |
2862 | } |
e14743d1 |
2863 | #endif |
2864 | if((sf->Rmask | sf->Gmask | sf->Bmask) == 0xffffff) |
2865 | { |
2866 | #if SDL_ALTIVEC_BLITTERS |
2867 | if(!(surface->map->dst->flags & SDL_HWSURFACE) |
2868 | && SDL_HasAltiVec()) |
2869 | return BlitRGBtoRGBSurfaceAlphaAltivec; |
2870 | #endif |
2871 | return BlitRGBtoRGBSurfaceAlpha; |
2872 | } |
2873 | } |
c85a5291 |
2874 | #ifdef __ARM_NEON__ |
2875 | if (sf->Gmask == df->Gmask && sf->Rmask == df->Bmask && sf->Bmask == df->Rmask |
2876 | && sf->Rshift % 8 == 0 && sf->Gshift % 8 == 0 && sf->Bshift % 8 == 0) |
2877 | { |
2878 | return BlitABGRtoXRGBalphaS_neon; |
2879 | } |
2880 | #endif |
e14743d1 |
2881 | #if SDL_ALTIVEC_BLITTERS |
2882 | if((sf->BytesPerPixel == 4) && |
2883 | !(surface->map->dst->flags & SDL_HWSURFACE) && SDL_HasAltiVec()) |
2884 | return Blit32to32SurfaceAlphaAltivec; |
2885 | else |
2886 | #endif |
2887 | return BlitNtoNSurfaceAlpha; |
2888 | |
2889 | case 3: |
2890 | default: |
2891 | return BlitNtoNSurfaceAlpha; |
2892 | } |
2893 | } |
2894 | } else { |
2895 | /* Per-pixel alpha blits */ |
2896 | switch(df->BytesPerPixel) { |
2897 | case 1: |
2898 | return BlitNto1PixelAlpha; |
2899 | |
2900 | case 2: |
2901 | #if SDL_ALTIVEC_BLITTERS |
2902 | if(sf->BytesPerPixel == 4 && !(surface->map->dst->flags & SDL_HWSURFACE) && |
2903 | df->Gmask == 0x7e0 && |
2904 | df->Bmask == 0x1f && SDL_HasAltiVec()) |
2905 | return Blit32to565PixelAlphaAltivec; |
2906 | else |
2907 | #endif |
2908 | if(sf->BytesPerPixel == 4 && sf->Amask == 0xff000000 |
2909 | && sf->Gmask == 0xff00 |
2910 | && ((sf->Rmask == 0xff && df->Rmask == 0x1f) |
2911 | || (sf->Bmask == 0xff && df->Bmask == 0x1f))) { |
2912 | if(df->Gmask == 0x7e0) |
2913 | return BlitARGBto565PixelAlpha; |
2914 | else if(df->Gmask == 0x3e0) |
2915 | return BlitARGBto555PixelAlpha; |
2916 | } |
2917 | return BlitNtoNPixelAlpha; |
2918 | |
2919 | case 4: |
2920 | if(sf->Rmask == df->Rmask |
2921 | && sf->Gmask == df->Gmask |
2922 | && sf->Bmask == df->Bmask |
2923 | && sf->BytesPerPixel == 4) |
2924 | { |
2925 | #if MMX_ASMBLIT |
2926 | if(sf->Rshift % 8 == 0 |
2927 | && sf->Gshift % 8 == 0 |
2928 | && sf->Bshift % 8 == 0 |
2929 | && sf->Ashift % 8 == 0 |
2930 | && sf->Aloss == 0) |
2931 | { |
2932 | if(SDL_Has3DNow()) |
2933 | return BlitRGBtoRGBPixelAlphaMMX3DNOW; |
2934 | if(SDL_HasMMX()) |
2935 | return BlitRGBtoRGBPixelAlphaMMX; |
2936 | } |
c85a5291 |
2937 | #endif |
2938 | #ifdef __ARM_NEON__ |
2939 | if(sf->Rshift % 8 == 0 |
2940 | && sf->Gshift % 8 == 0 |
2941 | && sf->Bshift % 8 == 0 |
2942 | && sf->Ashift % 8 == 0) |
2943 | { |
2944 | return BlitARGBtoXRGBalpha_neon; |
2945 | } |
e14743d1 |
2946 | #endif |
2947 | if(sf->Amask == 0xff000000) |
2948 | { |
2949 | #if SDL_ALTIVEC_BLITTERS |
2950 | if(!(surface->map->dst->flags & SDL_HWSURFACE) |
2951 | && SDL_HasAltiVec()) |
2952 | return BlitRGBtoRGBPixelAlphaAltivec; |
2953 | #endif |
2954 | return BlitRGBtoRGBPixelAlpha; |
2955 | } |
2956 | } |
a1f34081 |
2957 | #ifdef __ARM_NEON__ |
c85a5291 |
2958 | if (sf->Gmask == df->Gmask && sf->Rmask == df->Bmask && sf->Bmask == df->Rmask |
2959 | && sf->Rshift % 8 == 0 && sf->Gshift % 8 == 0 && sf->Bshift % 8 == 0 |
2960 | && sf->Amask == 0xff000000) |
a1f34081 |
2961 | { |
2962 | return BlitABGRtoXRGBalpha_neon; |
2963 | } |
2964 | #endif |
e14743d1 |
2965 | #if SDL_ALTIVEC_BLITTERS |
2966 | if (sf->Amask && sf->BytesPerPixel == 4 && |
2967 | !(surface->map->dst->flags & SDL_HWSURFACE) && SDL_HasAltiVec()) |
2968 | return Blit32to32PixelAlphaAltivec; |
2969 | else |
2970 | #endif |
2971 | return BlitNtoNPixelAlpha; |
2972 | |
2973 | case 3: |
2974 | default: |
2975 | return BlitNtoNPixelAlpha; |
2976 | } |
2977 | } |
2978 | } |
2979 | |