This repository has been archived by the owner on Feb 11, 2021. It is now read-only.
SDL_blit_A.c
1
2
/*
SDL - Simple DirectMedia Layer
3
Copyright (C) 1997-2006 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
11
12
This library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13
Lesser General Public License for more details.
15
16
17
You should have received a copy of the GNU Lesser General Public
License along with this library; if not, write to the Free Software
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
25
26
#include "SDL_video.h"
#include "SDL_blit.h"
27
28
#if SDL_ASSEMBLY_ROUTINES
#if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))
29
#define MMX_ASMBLIT 1
30
31
32
33
#define GCC_ASMBLIT 1
#elif defined(_MSC_VER) && (_MSC_VER >= 1200) && defined(_M_IX86)
#define MMX_ASMBLIT 1
#define MSVC_ASMBLIT 1
35
#endif /* SDL_ASSEMBLY_ROUTINES */
37
/* Function to check the CPU flags */
38
#include "SDL_cpuinfo.h"
39
#if GCC_ASMBLIT
40
#include "mmx.h"
41
42
43
#elif MSVC_ASMBLIT
#include <mmintrin.h>
#include <mm3dnow.h>
44
45
#endif
46
47
48
/* Functions to perform alpha blended blitting */
/* N->1 blending with per-surface alpha */
50
BlitNto1SurfaceAlpha(SDL_BlitInfo * info)
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
int width = info->d_width;
int height = info->d_height;
Uint8 *src = info->s_pixels;
int srcskip = info->s_skip;
Uint8 *dst = info->d_pixels;
int dstskip = info->d_skip;
Uint8 *palmap = info->table;
SDL_PixelFormat *srcfmt = info->src;
SDL_PixelFormat *dstfmt = info->dst;
int srcbpp = srcfmt->BytesPerPixel;
const unsigned A = srcfmt->alpha;
while (height--) {
/* *INDENT-OFF* */
67
68
DUFFS_LOOP4(
{
70
71
72
73
74
75
unsigned sR;
unsigned sG;
unsigned sB;
unsigned dR;
unsigned dG;
unsigned dB;
76
DISEMBLE_RGB(src, srcbpp, srcfmt, Pixel, sR, sG, sB);
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
dR = dstfmt->palette->colors[*dst].r;
dG = dstfmt->palette->colors[*dst].g;
dB = dstfmt->palette->colors[*dst].b;
ALPHA_BLEND(sR, sG, sB, A, dR, dG, dB);
dR &= 0xff;
dG &= 0xff;
dB &= 0xff;
/* Pack RGB into 8bit pixel */
if ( palmap == NULL ) {
*dst =((dR>>5)<<(3+2))|
((dG>>5)<<(2))|
((dB>>6)<<(0));
} else {
*dst = palmap[((dR>>5)<<(3+2))|
((dG>>5)<<(2)) |
((dB>>6)<<(0))];
}
dst++;
src += srcbpp;
},
width);
98
99
100
101
/* *INDENT-ON* */
src += srcskip;
dst += dstskip;
}
102
103
104
}
/* N->1 blending with pixel alpha */
106
BlitNto1PixelAlpha(SDL_BlitInfo * info)
107
{
108
109
110
111
112
113
114
115
116
117
118
119
120
121
int width = info->d_width;
int height = info->d_height;
Uint8 *src = info->s_pixels;
int srcskip = info->s_skip;
Uint8 *dst = info->d_pixels;
int dstskip = info->d_skip;
Uint8 *palmap = info->table;
SDL_PixelFormat *srcfmt = info->src;
SDL_PixelFormat *dstfmt = info->dst;
int srcbpp = srcfmt->BytesPerPixel;
/* FIXME: fix alpha bit field expansion here too? */
while (height--) {
/* *INDENT-OFF* */
122
123
DUFFS_LOOP4(
{
124
Uint32 Pixel;
125
126
127
128
129
130
131
unsigned sR;
unsigned sG;
unsigned sB;
unsigned sA;
unsigned dR;
unsigned dG;
unsigned dB;
132
DISEMBLE_RGBA(src,srcbpp,srcfmt,Pixel,sR,sG,sB,sA);
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
dR = dstfmt->palette->colors[*dst].r;
dG = dstfmt->palette->colors[*dst].g;
dB = dstfmt->palette->colors[*dst].b;
ALPHA_BLEND(sR, sG, sB, sA, dR, dG, dB);
dR &= 0xff;
dG &= 0xff;
dB &= 0xff;
/* Pack RGB into 8bit pixel */
if ( palmap == NULL ) {
*dst =((dR>>5)<<(3+2))|
((dG>>5)<<(2))|
((dB>>6)<<(0));
} else {
*dst = palmap[((dR>>5)<<(3+2))|
((dG>>5)<<(2)) |
((dB>>6)<<(0)) ];
}
dst++;
src += srcbpp;
},
width);
154
155
156
157
/* *INDENT-ON* */
src += srcskip;
dst += dstskip;
}
158
159
160
}
/* colorkeyed N->1 blending with per-surface alpha */
162
BlitNto1SurfaceAlphaKey(SDL_BlitInfo * info)
163
{
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
int width = info->d_width;
int height = info->d_height;
Uint8 *src = info->s_pixels;
int srcskip = info->s_skip;
Uint8 *dst = info->d_pixels;
int dstskip = info->d_skip;
Uint8 *palmap = info->table;
SDL_PixelFormat *srcfmt = info->src;
SDL_PixelFormat *dstfmt = info->dst;
int srcbpp = srcfmt->BytesPerPixel;
Uint32 ckey = srcfmt->colorkey;
const int A = srcfmt->alpha;
while (height--) {
/* *INDENT-OFF* */
180
181
DUFFS_LOOP(
{
182
Uint32 Pixel;
183
184
185
186
187
188
unsigned sR;
unsigned sG;
unsigned sB;
unsigned dR;
unsigned dG;
unsigned dB;
189
190
DISEMBLE_RGB(src, srcbpp, srcfmt, Pixel, sR, sG, sB);
if ( Pixel != ckey ) {
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
dR = dstfmt->palette->colors[*dst].r;
dG = dstfmt->palette->colors[*dst].g;
dB = dstfmt->palette->colors[*dst].b;
ALPHA_BLEND(sR, sG, sB, A, dR, dG, dB);
dR &= 0xff;
dG &= 0xff;
dB &= 0xff;
/* Pack RGB into 8bit pixel */
if ( palmap == NULL ) {
*dst =((dR>>5)<<(3+2))|
((dG>>5)<<(2)) |
((dB>>6)<<(0));
} else {
*dst = palmap[((dR>>5)<<(3+2))|
((dG>>5)<<(2)) |
((dB>>6)<<(0)) ];
}
}
dst++;
src += srcbpp;
},
width);
213
214
215
216
/* *INDENT-ON* */
src += srcskip;
dst += dstskip;
}
217
218
}
219
#if GCC_ASMBLIT
220
/* fast RGB888->(A)RGB888 blending with surface alpha=128 special case */
222
BlitRGBtoRGBSurfaceAlpha128MMX(SDL_BlitInfo * info)
224
225
226
227
228
229
230
231
232
233
int width = info->d_width;
int height = info->d_height;
Uint32 *srcp = (Uint32 *) info->s_pixels;
int srcskip = info->s_skip >> 2;
Uint32 *dstp = (Uint32 *) info->d_pixels;
int dstskip = info->d_skip >> 2;
Uint32 dalpha = info->dst->Amask;
Uint8 load[8];
*(Uint64 *) load = 0x00fefefe00fefefeULL; /* alpha128 mask */
234
movq_m2r(*load, mm4); /* alpha128 mask -> mm4 */
235
*(Uint64 *) load = 0x0001010100010101ULL; /* !alpha128 mask */
236
237
238
movq_m2r(*load, mm3); /* !alpha128 mask -> mm3 */
movd_m2r(dalpha, mm7); /* dst alpha mask */
punpckldq_r2r(mm7, mm7); /* dst alpha mask | dst alpha mask -> mm7 */
239
240
while (height--) {
/* *INDENT-OFF* */
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
DUFFS_LOOP_DOUBLE2(
{
Uint32 s = *srcp++;
Uint32 d = *dstp;
*dstp++ = ((((s & 0x00fefefe) + (d & 0x00fefefe)) >> 1)
+ (s & d & 0x00010101)) | dalpha;
},{
movq_m2r((*dstp), mm2);/* 2 x dst -> mm2(ARGBARGB) */
movq_r2r(mm2, mm6); /* 2 x dst -> mm6(ARGBARGB) */
movq_m2r((*srcp), mm1);/* 2 x src -> mm1(ARGBARGB) */
movq_r2r(mm1, mm5); /* 2 x src -> mm5(ARGBARGB) */
pand_r2r(mm4, mm6); /* dst & mask -> mm6 */
pand_r2r(mm4, mm5); /* src & mask -> mm5 */
paddd_r2r(mm6, mm5); /* mm6 + mm5 -> mm5 */
pand_r2r(mm1, mm2); /* src & dst -> mm2 */
psrld_i2r(1, mm5); /* mm5 >> 1 -> mm5 */
pand_r2r(mm3, mm2); /* mm2 & !mask -> mm2 */
paddd_r2r(mm5, mm2); /* mm5 + mm2 -> mm2 */
por_r2r(mm7, mm2); /* mm7(full alpha) | mm2 -> mm2 */
movq_r2m(mm2, (*dstp));/* mm2 -> 2 x dst pixels */
dstp += 2;
srcp += 2;
}, width);
267
268
269
270
/* *INDENT-ON* */
srcp += srcskip;
dstp += dstskip;
}
271
emms();
272
273
274
}
/* fast RGB888->(A)RGB888 blending with surface alpha */
276
BlitRGBtoRGBSurfaceAlphaMMX(SDL_BlitInfo * info)
278
279
280
281
282
SDL_PixelFormat *df = info->dst;
unsigned alpha = info->src->alpha;
if (alpha == 128 && (df->Rmask | df->Gmask | df->Bmask) == 0x00FFFFFF) {
/* only call a128 version when R,G,B occupy lower bits */
283
BlitRGBtoRGBSurfaceAlpha128MMX(info);
284
285
286
287
288
289
290
291
} else {
int width = info->d_width;
int height = info->d_height;
Uint32 *srcp = (Uint32 *) info->s_pixels;
int srcskip = info->s_skip >> 2;
Uint32 *dstp = (Uint32 *) info->d_pixels;
int dstskip = info->d_skip >> 2;
292
pxor_r2r(mm5, mm5); /* 0 -> mm5 */
293
/* form the alpha mult */
294
295
296
movd_m2r(alpha, mm4); /* 0000000A -> mm4 */
punpcklwd_r2r(mm4, mm4); /* 00000A0A -> mm4 */
punpckldq_r2r(mm4, mm4); /* 0A0A0A0A -> mm4 */
297
298
299
alpha =
(0xff << df->Rshift) | (0xff << df->Gshift) | (0xff << df->
Bshift);
300
301
302
movd_m2r(alpha, mm0); /* 00000FFF -> mm0 */
punpcklbw_r2r(mm0, mm0); /* 00FFFFFF -> mm0 */
pand_r2r(mm0, mm4); /* 0A0A0A0A -> mm4, minus 1 chan */
303
/* at this point mm4 can be 000A0A0A or 0A0A0A00 or another combo */
304
305
movd_m2r(df->Amask, mm7); /* dst alpha mask */
punpckldq_r2r(mm7, mm7); /* dst alpha mask | dst alpha mask -> mm7 */
306
307
308
while (height--) {
/* *INDENT-OFF* */
309
310
DUFFS_LOOP_DOUBLE2({
/* One Pixel Blend */
311
312
313
314
315
316
317
318
319
320
321
322
323
movd_m2r((*srcp), mm1);/* src(ARGB) -> mm1 (0000ARGB)*/
movd_m2r((*dstp), mm2);/* dst(ARGB) -> mm2 (0000ARGB)*/
punpcklbw_r2r(mm5, mm1); /* 0A0R0G0B -> mm1(src) */
punpcklbw_r2r(mm5, mm2); /* 0A0R0G0B -> mm2(dst) */
psubw_r2r(mm2, mm1);/* src - dst -> mm1 */
pmullw_r2r(mm4, mm1); /* mm1 * alpha -> mm1 */
psrlw_i2r(8, mm1); /* mm1 >> 8 -> mm1 */
paddb_r2r(mm1, mm2); /* mm1 + mm2(dst) -> mm2 */
packuswb_r2r(mm5, mm2); /* ARGBARGB -> mm2 */
por_r2r(mm7, mm2); /* mm7(full alpha) | mm2 -> mm2 */
movd_r2m(mm2, *dstp);/* mm2 -> pixel */
324
325
326
++srcp;
++dstp;
},{
327
/* Two Pixels Blend */
328
movq_m2r((*srcp), mm0);/* 2 x src -> mm0(ARGBARGB)*/
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
movq_m2r((*dstp), mm2);/* 2 x dst -> mm2(ARGBARGB) */
movq_r2r(mm0, mm1); /* 2 x src -> mm1(ARGBARGB) */
movq_r2r(mm2, mm6); /* 2 x dst -> mm6(ARGBARGB) */
punpcklbw_r2r(mm5, mm0); /* low - 0A0R0G0B -> mm0(src1) */
punpckhbw_r2r(mm5, mm1); /* high - 0A0R0G0B -> mm1(src2) */
punpcklbw_r2r(mm5, mm2); /* low - 0A0R0G0B -> mm2(dst1) */
punpckhbw_r2r(mm5, mm6); /* high - 0A0R0G0B -> mm6(dst2) */
psubw_r2r(mm2, mm0);/* src1 - dst1 -> mm0 */
pmullw_r2r(mm4, mm0); /* mm0 * alpha -> mm0 */
psrlw_i2r(8, mm0); /* mm0 >> 8 -> mm1 */
paddb_r2r(mm0, mm2); /* mm0 + mm2(dst1) -> mm2 */
psubw_r2r(mm6, mm1);/* src2 - dst2 -> mm1 */
pmullw_r2r(mm4, mm1); /* mm1 * alpha -> mm1 */
psrlw_i2r(8, mm1); /* mm1 >> 8 -> mm1 */
paddb_r2r(mm1, mm6); /* mm1 + mm6(dst2) -> mm6 */
packuswb_r2r(mm6, mm2); /* ARGBARGB -> mm2 */
por_r2r(mm7, mm2); /* mm7(dst alpha) | mm2 -> mm2 */
movq_r2m(mm2, *dstp);/* mm2 -> 2 x pixel */
srcp += 2;
dstp += 2;
}, width);
356
357
358
359
/* *INDENT-ON* */
srcp += srcskip;
dstp += dstskip;
}
360
emms();
362
363
364
}
/* fast ARGB888->(A)RGB888 blending with pixel alpha */
366
BlitRGBtoRGBPixelAlphaMMX(SDL_BlitInfo * info)
368
369
370
371
372
373
374
375
376
int width = info->d_width;
int height = info->d_height;
Uint32 *srcp = (Uint32 *) info->s_pixels;
int srcskip = info->s_skip >> 2;
Uint32 *dstp = (Uint32 *) info->d_pixels;
int dstskip = info->d_skip >> 2;
SDL_PixelFormat *sf = info->src;
Uint32 amask = sf->Amask;
377
pxor_r2r(mm6, mm6); /* 0 -> mm6 */
378
/* form multiplication mask */
379
380
381
382
383
movd_m2r(sf->Amask, mm7); /* 0000F000 -> mm7 */
punpcklbw_r2r(mm7, mm7); /* FF000000 -> mm7 */
pcmpeqb_r2r(mm0, mm0); /* FFFFFFFF -> mm0 */
movq_r2r(mm0, mm3); /* FFFFFFFF -> mm3 (for later) */
pxor_r2r(mm0, mm7); /* 00FFFFFF -> mm7 (mult mask) */
384
/* form channel masks */
385
386
387
388
movq_r2r(mm7, mm0); /* 00FFFFFF -> mm0 */
packsswb_r2r(mm6, mm0); /* 00000FFF -> mm0 (channel mask) */
packsswb_r2r(mm6, mm3); /* 0000FFFF -> mm3 */
pxor_r2r(mm0, mm3); /* 0000F000 -> mm3 (~channel mask) */
389
/* get alpha channel shift */
390
movd_m2r(sf->Ashift, mm5); /* Ashift -> mm5 */
391
392
393
while (height--) {
/* *INDENT-OFF* */
394
DUFFS_LOOP4({
395
Uint32 alpha = *srcp & amask;
396
/* FIXME: Here we special-case opaque alpha since the
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
compositioning used (>>8 instead of /255) doesn't handle
it correctly. Also special-case alpha=0 for speed?
Benchmark this! */
if(alpha == 0) {
/* do nothing */
} else if(alpha == amask) {
/* opaque alpha -- copy RGB, keep dst alpha */
/* using MMX here to free up regular registers for other things */
movd_m2r((*srcp), mm1);/* src(ARGB) -> mm1 (0000ARGB)*/
movd_m2r((*dstp), mm2);/* dst(ARGB) -> mm2 (0000ARGB)*/
pand_r2r(mm0, mm1); /* src & chanmask -> mm1 */
pand_r2r(mm3, mm2); /* dst & ~chanmask -> mm2 */
por_r2r(mm1, mm2); /* src | dst -> mm2 */
movd_r2m(mm2, (*dstp)); /* mm2 -> dst */
} else {
movd_m2r((*srcp), mm1);/* src(ARGB) -> mm1 (0000ARGB)*/
punpcklbw_r2r(mm6, mm1); /* 0A0R0G0B -> mm1 */
movd_m2r((*dstp), mm2);/* dst(ARGB) -> mm2 (0000ARGB)*/
punpcklbw_r2r(mm6, mm2); /* 0A0R0G0B -> mm2 */
__asm__ __volatile__ (
"movd %0, %%mm4"
: : "r" (alpha) ); /* 0000A000 -> mm4 */
psrld_r2r(mm5, mm4); /* mm4 >> mm5 -> mm4 (0000000A) */
punpcklwd_r2r(mm4, mm4); /* 00000A0A -> mm4 */
punpcklwd_r2r(mm4, mm4); /* 0A0A0A0A -> mm4 */
pand_r2r(mm7, mm4); /* 000A0A0A -> mm4, preserve dst alpha on add */
/* blend */
psubw_r2r(mm2, mm1);/* src - dst -> mm1 */
pmullw_r2r(mm4, mm1); /* mm1 * alpha -> mm1 */
psrlw_i2r(8, mm1); /* mm1 >> 8 -> mm1(000R0G0B) */
paddb_r2r(mm1, mm2); /* mm1 + mm2(dst) -> mm2 */
packuswb_r2r(mm6, mm2); /* 0000ARGB -> mm2 */
movd_r2m(mm2, *dstp);/* mm2 -> dst */
434
435
436
437
}
++srcp;
++dstp;
}, width);
438
439
440
441
/* *INDENT-ON* */
srcp += srcskip;
dstp += dstskip;
}
442
emms();
445
446
447
448
/* End GCC_ASMBLIT */
#elif MSVC_ASMBLIT
/* fast RGB888->(A)RGB888 blending with surface alpha=128 special case */
450
BlitRGBtoRGBSurfaceAlpha128MMX(SDL_BlitInfo * info)
451
{
452
453
454
455
456
457
458
459
460
461
int width = info->d_width;
int height = info->d_height;
Uint32 *srcp = (Uint32 *) info->s_pixels;
int srcskip = info->s_skip >> 2;
Uint32 *dstp = (Uint32 *) info->d_pixels;
int dstskip = info->d_skip >> 2;
Uint32 dalpha = info->dst->Amask;
__m64 src1, src2, dst1, dst2, lmask, hmask, dsta;
462
463
464
hmask = _mm_set_pi32(0x00fefefe, 0x00fefefe); /* alpha128 mask -> hmask */
lmask = _mm_set_pi32(0x00010101, 0x00010101); /* !alpha128 mask -> lmask */
dsta = _mm_set_pi32(dalpha, dalpha); /* dst alpha mask -> dsta */
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
while (height--) {
int n = width;
if (n & 1) {
Uint32 s = *srcp++;
Uint32 d = *dstp;
*dstp++ = ((((s & 0x00fefefe) + (d & 0x00fefefe)) >> 1)
+ (s & d & 0x00010101)) | dalpha;
n--;
}
for (n >>= 1; n > 0; --n) {
dst1 = *(__m64 *) dstp; /* 2 x dst -> dst1(ARGBARGB) */
dst2 = dst1; /* 2 x dst -> dst2(ARGBARGB) */
src1 = *(__m64 *) srcp; /* 2 x src -> src1(ARGBARGB) */
src2 = src1; /* 2 x src -> src2(ARGBARGB) */
483
484
485
486
dst2 = _mm_and_si64(dst2, hmask); /* dst & mask -> dst2 */
src2 = _mm_and_si64(src2, hmask); /* src & mask -> src2 */
src2 = _mm_add_pi32(src2, dst2); /* dst2 + src2 -> src2 */
src2 = _mm_srli_pi32(src2, 1); /* src2 >> 1 -> src2 */
488
489
490
491
dst1 = _mm_and_si64(dst1, src1); /* src & dst -> dst1 */
dst1 = _mm_and_si64(dst1, lmask); /* dst1 & !mask -> dst1 */
dst1 = _mm_add_pi32(dst1, src2); /* src2 + dst1 -> dst1 */
dst1 = _mm_or_si64(dst1, dsta); /* dsta(full alpha) | dst1 -> dst1 */
492
493
494
495
496
497
498
499
500
*(__m64 *) dstp = dst1; /* dst1 -> 2 x dst pixels */
dstp += 2;
srcp += 2;
}
srcp += srcskip;
dstp += dstskip;
}
501
_mm_empty();
502
503
504
}
/* fast RGB888->(A)RGB888 blending with surface alpha */
506
BlitRGBtoRGBSurfaceAlphaMMX(SDL_BlitInfo * info)
507
{
508
509
510
SDL_PixelFormat *df = info->dst;
Uint32 chanmask = df->Rmask | df->Gmask | df->Bmask;
unsigned alpha = info->src->alpha;
512
513
if (alpha == 128 && (df->Rmask | df->Gmask | df->Bmask) == 0x00FFFFFF) {
/* only call a128 version when R,G,B occupy lower bits */
514
BlitRGBtoRGBSurfaceAlpha128MMX(info);
515
516
517
518
519
520
521
522
523
524
525
526
} else {
int width = info->d_width;
int height = info->d_height;
Uint32 *srcp = (Uint32 *) info->s_pixels;
int srcskip = info->s_skip >> 2;
Uint32 *dstp = (Uint32 *) info->d_pixels;
int dstskip = info->d_skip >> 2;
Uint32 dalpha = df->Amask;
Uint32 amult;
__m64 src1, src2, dst1, dst2, mm_alpha, mm_zero, dsta;
527
mm_zero = _mm_setzero_si64(); /* 0 -> mm_zero */
528
529
530
531
532
533
/* form the alpha mult */
amult = alpha | (alpha << 8);
amult = amult | (amult << 16);
chanmask =
(0xff << df->Rshift) | (0xff << df->Gshift) | (0xff << df->
Bshift);
534
535
mm_alpha = _mm_set_pi32(0, amult & chanmask); /* 0000AAAA -> mm_alpha, minus 1 chan */
mm_alpha = _mm_unpacklo_pi8(mm_alpha, mm_zero); /* 0A0A0A0A -> mm_alpha, minus 1 chan */
536
/* at this point mm_alpha can be 000A0A0A or 0A0A0A00 or another combo */
537
dsta = _mm_set_pi32(dalpha, dalpha); /* dst alpha mask -> dsta */
538
539
540
541
542
while (height--) {
int n = width;
if (n & 1) {
/* One Pixel Blend */
543
544
src2 = _mm_cvtsi32_si64(*srcp); /* src(ARGB) -> src2 (0000ARGB) */
src2 = _mm_unpacklo_pi8(src2, mm_zero); /* 0A0R0G0B -> src2 */
546
547
dst1 = _mm_cvtsi32_si64(*dstp); /* dst(ARGB) -> dst1 (0000ARGB) */
dst1 = _mm_unpacklo_pi8(dst1, mm_zero); /* 0A0R0G0B -> dst1 */
549
550
551
552
src2 = _mm_sub_pi16(src2, dst1); /* src2 - dst2 -> src2 */
src2 = _mm_mullo_pi16(src2, mm_alpha); /* src2 * alpha -> src2 */
src2 = _mm_srli_pi16(src2, 8); /* src2 >> 8 -> src2 */
dst1 = _mm_add_pi8(src2, dst1); /* src2 + dst1 -> dst1 */
554
555
556
dst1 = _mm_packs_pu16(dst1, mm_zero); /* 0000ARGB -> dst1 */
dst1 = _mm_or_si64(dst1, dsta); /* dsta | dst1 -> dst1 */
*dstp = _mm_cvtsi64_si32(dst1); /* dst1 -> pixel */
557
558
559
560
561
562
++srcp;
++dstp;
n--;
}
564
565
566
567
for (n >>= 1; n > 0; --n) {
/* Two Pixels Blend */
src1 = *(__m64 *) srcp; /* 2 x src -> src1(ARGBARGB) */
src2 = src1; /* 2 x src -> src2(ARGBARGB) */
568
569
src1 = _mm_unpacklo_pi8(src1, mm_zero); /* low - 0A0R0G0B -> src1 */
src2 = _mm_unpackhi_pi8(src2, mm_zero); /* high - 0A0R0G0B -> src2 */
571
572
dst1 = *(__m64 *) dstp; /* 2 x dst -> dst1(ARGBARGB) */
dst2 = dst1; /* 2 x dst -> dst2(ARGBARGB) */
573
574
dst1 = _mm_unpacklo_pi8(dst1, mm_zero); /* low - 0A0R0G0B -> dst1 */
dst2 = _mm_unpackhi_pi8(dst2, mm_zero); /* high - 0A0R0G0B -> dst2 */
576
577
578
579
src1 = _mm_sub_pi16(src1, dst1); /* src1 - dst1 -> src1 */
src1 = _mm_mullo_pi16(src1, mm_alpha); /* src1 * alpha -> src1 */
src1 = _mm_srli_pi16(src1, 8); /* src1 >> 8 -> src1 */
dst1 = _mm_add_pi8(src1, dst1); /* src1 + dst1(dst1) -> dst1 */
581
582
583
584
src2 = _mm_sub_pi16(src2, dst2); /* src2 - dst2 -> src2 */
src2 = _mm_mullo_pi16(src2, mm_alpha); /* src2 * alpha -> src2 */
src2 = _mm_srli_pi16(src2, 8); /* src2 >> 8 -> src2 */
dst2 = _mm_add_pi8(src2, dst2); /* src2 + dst2(dst2) -> dst2 */
586
587
dst1 = _mm_packs_pu16(dst1, dst2); /* 0A0R0G0B(res1), 0A0R0G0B(res2) -> dst1(ARGBARGB) */
dst1 = _mm_or_si64(dst1, dsta); /* dsta | dst1 -> dst1 */
588
589
590
591
592
593
594
595
596
*(__m64 *) dstp = dst1; /* dst1 -> 2 x pixel */
srcp += 2;
dstp += 2;
}
srcp += srcskip;
dstp += dstskip;
}
597
_mm_empty();
599
600
601
}
/* fast ARGB888->(A)RGB888 blending with pixel alpha */
603
BlitRGBtoRGBPixelAlphaMMX(SDL_BlitInfo * info)
604
{
605
606
607
608
609
610
611
612
613
614
615
616
617
618
int width = info->d_width;
int height = info->d_height;
Uint32 *srcp = (Uint32 *) info->s_pixels;
int srcskip = info->s_skip >> 2;
Uint32 *dstp = (Uint32 *) info->d_pixels;
int dstskip = info->d_skip >> 2;
SDL_PixelFormat *sf = info->src;
Uint32 chanmask = sf->Rmask | sf->Gmask | sf->Bmask;
Uint32 amask = sf->Amask;
Uint32 ashift = sf->Ashift;
Uint64 multmask;
__m64 src1, dst1, mm_alpha, mm_zero, dmask;
619
mm_zero = _mm_setzero_si64(); /* 0 -> mm_zero */
620
621
622
/* *INDENT-OFF* */
multmask = ~(0xFFFFI64 << (ashift * 2));
/* *INDENT-ON* */
623
624
625
626
dmask = *(__m64 *) & multmask; /* dst alpha mask -> dmask */
while (height--) {
/* *INDENT-OFF* */
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
DUFFS_LOOP4({
Uint32 alpha = *srcp & amask;
if (alpha == 0) {
/* do nothing */
} else if (alpha == amask) {
/* opaque alpha -- copy RGB, keep dst alpha */
*dstp = (*srcp & chanmask) | (*dstp & ~chanmask);
} else {
src1 = _mm_cvtsi32_si64(*srcp); /* src(ARGB) -> src1 (0000ARGB)*/
src1 = _mm_unpacklo_pi8(src1, mm_zero); /* 0A0R0G0B -> src1 */
dst1 = _mm_cvtsi32_si64(*dstp); /* dst(ARGB) -> dst1 (0000ARGB)*/
dst1 = _mm_unpacklo_pi8(dst1, mm_zero); /* 0A0R0G0B -> dst1 */
mm_alpha = _mm_cvtsi32_si64(alpha); /* alpha -> mm_alpha (0000000A) */
mm_alpha = _mm_srli_si64(mm_alpha, ashift); /* mm_alpha >> ashift -> mm_alpha(0000000A) */
mm_alpha = _mm_unpacklo_pi16(mm_alpha, mm_alpha); /* 00000A0A -> mm_alpha */
mm_alpha = _mm_unpacklo_pi32(mm_alpha, mm_alpha); /* 0A0A0A0A -> mm_alpha */
mm_alpha = _mm_and_si64(mm_alpha, dmask); /* 000A0A0A -> mm_alpha, preserve dst alpha on add */
/* blend */
src1 = _mm_sub_pi16(src1, dst1);/* src1 - dst1 -> src1 */
src1 = _mm_mullo_pi16(src1, mm_alpha); /* (src1 - dst1) * alpha -> src1 */
src1 = _mm_srli_pi16(src1, 8); /* src1 >> 8 -> src1(000R0G0B) */
dst1 = _mm_add_pi8(src1, dst1); /* src1 + dst1 -> dst1(0A0R0G0B) */
dst1 = _mm_packs_pu16(dst1, mm_zero); /* 0000ARGB -> dst1 */
*dstp = _mm_cvtsi64_si32(dst1); /* dst1 -> pixel */
}
++srcp;
++dstp;
}, width);
659
660
661
662
/* *INDENT-ON* */
srcp += srcskip;
dstp += dstskip;
}
663
_mm_empty();
664
}
666
667
668
/* End MSVC_ASMBLIT */
#endif /* GCC_ASMBLIT, MSVC_ASMBLIT */
670
#if SDL_ALTIVEC_BLITTERS
671
672
673
#if __MWERKS__
#pragma altivec_model on
#endif
674
#if HAVE_ALTIVEC_H
675
#include <altivec.h>
676
#endif
677
#include <assert.h>
679
#if (defined(__MACOSX__) && (__GNUC__ < 4))
680
#define VECUINT8_LITERAL(a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p) \
681
(vector unsigned char) ( a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p )
682
#define VECUINT16_LITERAL(a,b,c,d,e,f,g,h) \
683
684
(vector unsigned short) ( a,b,c,d,e,f,g,h )
#else
685
#define VECUINT8_LITERAL(a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p) \
686
(vector unsigned char) { a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p }
687
#define VECUINT16_LITERAL(a,b,c,d,e,f,g,h) \
688
689
690
(vector unsigned short) { a,b,c,d,e,f,g,h }
#endif
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
#define UNALIGNED_PTR(x) (((size_t) x) & 0x0000000F)
#define VECPRINT(msg, v) do { \
vector unsigned int tmpvec = (vector unsigned int)(v); \
unsigned int *vp = (unsigned int *)&tmpvec; \
printf("%s = %08X %08X %08X %08X\n", msg, vp[0], vp[1], vp[2], vp[3]); \
} while (0)
/* the permuation vector that takes the high bytes out of all the appropriate shorts
(vector unsigned char)(
0x00, 0x10, 0x02, 0x12,
0x04, 0x14, 0x06, 0x16,
0x08, 0x18, 0x0A, 0x1A,
0x0C, 0x1C, 0x0E, 0x1E );
*/
#define VEC_MERGE_PERMUTE() (vec_add(vec_lvsl(0, (int*)NULL), (vector unsigned char)vec_splat_u16(0x0F)))
#define VEC_U32_24() (vec_add(vec_splat_u32(12), vec_splat_u32(12)))
#define VEC_ALPHA_MASK() ((vector unsigned char)vec_sl((vector unsigned int)vec_splat_s8(-1), VEC_U32_24()))
#define VEC_ALIGNER(src) ((UNALIGNED_PTR(src)) \
? vec_lvsl(0, src) \
: vec_add(vec_lvsl(8, src), vec_splat_u8(8)))
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
#define VEC_MULTIPLY_ALPHA(vs, vd, valpha, mergePermute, v1_16, v8_16) do { \
/* vtemp1 contains source AAGGAAGGAAGGAAGG */ \
vector unsigned short vtemp1 = vec_mule(vs, valpha); \
/* vtemp2 contains source RRBBRRBBRRBBRRBB */ \
vector unsigned short vtemp2 = vec_mulo(vs, valpha); \
/* valpha2 is 255-alpha */ \
vector unsigned char valpha2 = vec_nor(valpha, valpha); \
/* vtemp3 contains dest AAGGAAGGAAGGAAGG */ \
vector unsigned short vtemp3 = vec_mule(vd, valpha2); \
/* vtemp4 contains dest RRBBRRBBRRBBRRBB */ \
vector unsigned short vtemp4 = vec_mulo(vd, valpha2); \
/* add source and dest */ \
vtemp1 = vec_add(vtemp1, vtemp3); \
vtemp2 = vec_add(vtemp2, vtemp4); \
/* vtemp1 = (vtemp1 + 1) + ((vtemp1 + 1) >> 8) */ \
vtemp1 = vec_add(vtemp1, v1_16); \
vtemp3 = vec_sr(vtemp1, v8_16); \
vtemp1 = vec_add(vtemp1, vtemp3); \
/* vtemp2 = (vtemp2 + 1) + ((vtemp2 + 1) >> 8) */ \
vtemp2 = vec_add(vtemp2, v1_16); \
vtemp4 = vec_sr(vtemp2, v8_16); \
vtemp2 = vec_add(vtemp2, vtemp4); \
/* (>>8) and get ARGBARGBARGBARGB */ \
vd = (vector unsigned char)vec_perm(vtemp1, vtemp2, mergePermute); \
} while (0)
739
/* Calculate the permute vector used for 32->32 swizzling */
740
static vector unsigned char
741
calc_swizzle32(const SDL_PixelFormat * srcfmt, const SDL_PixelFormat * dstfmt)
742
743
744
745
746
747
748
749
750
751
752
753
{
/*
* We have to assume that the bits that aren't used by other
* colors is alpha, and it's one complete byte, since some formats
* leave alpha with a zero mask, but we should still swizzle the bits.
*/
/* ARGB */
const static struct SDL_PixelFormat default_pixel_format = {
NULL, 0, 0,
0, 0, 0, 0,
16, 8, 0, 24,
0x00FF0000, 0x0000FF00, 0x000000FF, 0xFF000000,
756
757
758
759
760
761
if (!srcfmt) {
srcfmt = &default_pixel_format;
}
if (!dstfmt) {
dstfmt = &default_pixel_format;
}
762
763
764
765
766
const vector unsigned char plus = VECUINT8_LITERAL(0x00, 0x00, 0x00, 0x00,
0x04, 0x04, 0x04, 0x04,
0x08, 0x08, 0x08, 0x08,
0x0C, 0x0C, 0x0C,
0x0C);
767
768
769
vector unsigned char vswiz;
vector unsigned int srcvec;
#define RESHIFT(X) (3 - ((X) >> 3))
770
771
772
Uint32 rmask = RESHIFT(srcfmt->Rshift) << (dstfmt->Rshift);
Uint32 gmask = RESHIFT(srcfmt->Gshift) << (dstfmt->Gshift);
Uint32 bmask = RESHIFT(srcfmt->Bshift) << (dstfmt->Bshift);
773
774
775
Uint32 amask;
/* Use zero for alpha if either surface doesn't have alpha */
if (dstfmt->Amask) {
777
778
((srcfmt->Amask) ? RESHIFT(srcfmt->Ashift) : 0x10) << (dstfmt->
Ashift);
779
} else {
780
781
782
amask =
0x10101010 & ((dstfmt->Rmask | dstfmt->Gmask | dstfmt->Bmask) ^
0xFFFFFFFF);
784
785
#undef RESHIFT
((unsigned int *) (char *) &srcvec)[0] = (rmask | gmask | bmask | amask);
786
vswiz = vec_add(plus, (vector unsigned char) vec_splat(srcvec, 0));
787
return (vswiz);
788
789
}
791
Blit32to565PixelAlphaAltivec(SDL_BlitInfo * info)
792
793
{
int height = info->d_height;
794
Uint8 *src = (Uint8 *) info->s_pixels;
795
int srcskip = info->s_skip;
796
Uint8 *dst = (Uint8 *) info->d_pixels;
797
798
799
int dstskip = info->d_skip;
SDL_PixelFormat *srcfmt = info->src;
800
801
802
803
804
805
806
vector unsigned char v0 = vec_splat_u8(0);
vector unsigned short v8_16 = vec_splat_u16(8);
vector unsigned short v1_16 = vec_splat_u16(1);
vector unsigned short v2_16 = vec_splat_u16(2);
vector unsigned short v3_16 = vec_splat_u16(3);
vector unsigned int v8_32 = vec_splat_u32(8);
vector unsigned int v16_32 = vec_add(v8_32, v8_32);
807
vector unsigned short v3f =
808
809
VECUINT16_LITERAL(0x003f, 0x003f, 0x003f, 0x003f,
0x003f, 0x003f, 0x003f, 0x003f);
810
vector unsigned short vfc =
811
812
VECUINT16_LITERAL(0x00fc, 0x00fc, 0x00fc, 0x00fc,
0x00fc, 0x00fc, 0x00fc, 0x00fc);
813
814
/*
815
816
817
818
0x10 - 0x1f is the alpha
0x00 - 0x0e evens are the red
0x01 - 0x0f odds are zero
*/
819
820
821
822
823
vector unsigned char vredalpha1 = VECUINT8_LITERAL(0x10, 0x00, 0x01, 0x01,
0x10, 0x02, 0x01, 0x01,
0x10, 0x04, 0x01, 0x01,
0x10, 0x06, 0x01,
0x01);
824
vector unsigned char vredalpha2 =
825
826
(vector unsigned char) (vec_add((vector unsigned int) vredalpha1,
vec_sl(v8_32, v16_32))
829
830
831
0x00 - 0x0f is ARxx ARxx ARxx ARxx
0x11 - 0x0f odds are blue
*/
832
833
834
835
vector unsigned char vblue1 = VECUINT8_LITERAL(0x00, 0x01, 0x02, 0x11,
0x04, 0x05, 0x06, 0x13,
0x08, 0x09, 0x0a, 0x15,
0x0c, 0x0d, 0x0e, 0x17);
836
vector unsigned char vblue2 =
837
(vector unsigned char) (vec_add((vector unsigned int) vblue1, v8_32)
840
841
842
0x00 - 0x0f is ARxB ARxB ARxB ARxB
0x10 - 0x0e evens are green
*/
843
844
845
846
vector unsigned char vgreen1 = VECUINT8_LITERAL(0x00, 0x01, 0x10, 0x03,
0x04, 0x05, 0x12, 0x07,
0x08, 0x09, 0x14, 0x0b,
0x0c, 0x0d, 0x16, 0x0f);
847
848
vector unsigned char vgreen2 =
(vector unsigned
849
char) (vec_add((vector unsigned int) vgreen1, vec_sl(v8_32, v8_32))
851
852
853
854
855
856
vector unsigned char vgmerge = VECUINT8_LITERAL(0x00, 0x02, 0x00, 0x06,
0x00, 0x0a, 0x00, 0x0e,
0x00, 0x12, 0x00, 0x16,
0x00, 0x1a, 0x00, 0x1e);
vector unsigned char mergePermute = VEC_MERGE_PERMUTE();
vector unsigned char vpermute = calc_swizzle32(srcfmt, NULL);
857
vector unsigned char valphaPermute =
858
vec_and(vec_lvsl(0, (int *) NULL), vec_splat_u8(0xC));
860
861
vector unsigned short vf800 = (vector unsigned short) vec_splat_u8(-7);
vf800 = vec_sl(vf800, vec_splat_u16(8));
862
863
while (height--) {
864
865
866
867
868
869
870
871
int extrawidth;
vector unsigned char valigner;
vector unsigned char vsrc;
vector unsigned char voverflow;
int width = info->d_width;
#define ONE_PIXEL_BLEND(condition, widthvar) \
while (condition) { \
872
Uint32 Pixel; \
873
unsigned sR, sG, sB, dR, dG, dB, sA; \
874
DISEMBLE_RGBA(src, 4, srcfmt, Pixel, sR, sG, sB, sA); \
875
876
877
878
879
880
881
882
883
884
885
886
887
888
if(sA) { \
unsigned short dstpixel = *((unsigned short *)dst); \
dR = (dstpixel >> 8) & 0xf8; \
dG = (dstpixel >> 3) & 0xfc; \
dB = (dstpixel << 3) & 0xf8; \
ACCURATE_ALPHA_BLEND(sR, sG, sB, sA, dR, dG, dB); \
*((unsigned short *)dst) = ( \
((dR & 0xf8) << 8) | ((dG & 0xfc) << 3) | (dB >> 3) \
); \
} \
src += 4; \
dst += 2; \
widthvar--; \
}
889
ONE_PIXEL_BLEND((UNALIGNED_PTR(dst)) && (width), width);
890
extrawidth = (width % 8);
891
892
valigner = VEC_ALIGNER(src);
vsrc = (vector unsigned char) vec_ld(0, src);
893
894
895
896
897
898
899
900
901
width -= extrawidth;
while (width) {
vector unsigned char valpha;
vector unsigned char vsrc1, vsrc2;
vector unsigned char vdst1, vdst2;
vector unsigned short vR, vG, vB;
vector unsigned short vpixel, vrpixel, vgpixel, vbpixel;
/* Load 8 pixels from src as ARGB */
902
903
904
voverflow = (vector unsigned char) vec_ld(15, src);
vsrc = vec_perm(vsrc, voverflow, valigner);
vsrc1 = vec_perm(vsrc, vsrc, vpermute);
905
src += 16;
906
907
908
vsrc = (vector unsigned char) vec_ld(15, src);
voverflow = vec_perm(voverflow, vsrc, valigner);
vsrc2 = vec_perm(voverflow, voverflow, vpermute);
909
910
911
src += 16;
/* Load 8 pixels from dst as XRGB */
912
913
914
915
voverflow = vec_ld(0, dst);
vR = vec_and((vector unsigned short) voverflow, vf800);
vB = vec_sl((vector unsigned short) voverflow, v3_16);
vG = vec_sl(vB, v2_16);
917
918
919
920
921
(vector unsigned char) vec_perm((vector unsigned char) vR,
(vector unsigned char) vR,
vredalpha1);
vdst1 = vec_perm(vdst1, (vector unsigned char) vB, vblue1);
vdst1 = vec_perm(vdst1, (vector unsigned char) vG, vgreen1);
923
924
925
926
927
(vector unsigned char) vec_perm((vector unsigned char) vR,
(vector unsigned char) vR,
vredalpha2);
vdst2 = vec_perm(vdst2, (vector unsigned char) vB, vblue2);
vdst2 = vec_perm(vdst2, (vector unsigned char) vG, vgreen2);
928
929
/* Alpha blend 8 pixels as ARGB */
930
931
932
933
934
935
valpha = vec_perm(vsrc1, v0, valphaPermute);
VEC_MULTIPLY_ALPHA(vsrc1, vdst1, valpha, mergePermute, v1_16,
v8_16);
valpha = vec_perm(vsrc2, v0, valphaPermute);
VEC_MULTIPLY_ALPHA(vsrc2, vdst2, valpha, mergePermute, v1_16,
v8_16);
936
937
/* Convert 8 pixels to 565 */
938
939
940
941
942
943
944
945
946
947
vpixel = (vector unsigned short) vec_packpx((vector unsigned int)
vdst1,
(vector unsigned int)
vdst2);
vgpixel = (vector unsigned short) vec_perm(vdst1, vdst2, vgmerge);
vgpixel = vec_and(vgpixel, vfc);
vgpixel = vec_sl(vgpixel, v3_16);
vrpixel = vec_sl(vpixel, v1_16);
vrpixel = vec_and(vrpixel, vf800);
vbpixel = vec_and(vpixel, v3f);
949
950
951
vec_or((vector unsigned char) vrpixel,
(vector unsigned char) vgpixel);
vdst1 = vec_or(vdst1, (vector unsigned char) vbpixel);
953
/* Store 8 pixels */
954
vec_st(vdst1, 0, dst);
955
956
957
958
width -= 8;
dst += 16;
}
959
ONE_PIXEL_BLEND((extrawidth), extrawidth);
960
961
962
963
964
965
#undef ONE_PIXEL_BLEND
src += srcskip;
dst += dstskip;
}
}
967
Blit32to32SurfaceAlphaKeyAltivec(SDL_BlitInfo * info)
968
969
970
{
unsigned alpha = info->src->alpha;
int height = info->d_height;
971
Uint32 *srcp = (Uint32 *) info->s_pixels;
972
int srcskip = info->s_skip >> 2;
973
Uint32 *dstp = (Uint32 *) info->d_pixels;
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
int dstskip = info->d_skip >> 2;
SDL_PixelFormat *srcfmt = info->src;
SDL_PixelFormat *dstfmt = info->dst;
unsigned sA = srcfmt->alpha;
unsigned dA = dstfmt->Amask ? SDL_ALPHA_OPAQUE : 0;
Uint32 rgbmask = srcfmt->Rmask | srcfmt->Gmask | srcfmt->Bmask;
Uint32 ckey = info->src->colorkey;
vector unsigned char mergePermute;
vector unsigned char vsrcPermute;
vector unsigned char vdstPermute;
vector unsigned char vsdstPermute;
vector unsigned char valpha;
vector unsigned char valphamask;
vector unsigned char vbits;
vector unsigned char v0;
vector unsigned short v1;
vector unsigned short v8;
vector unsigned int vckey;
vector unsigned int vrgbmask;
994
995
996
997
mergePermute = VEC_MERGE_PERMUTE();
v0 = vec_splat_u8(0);
v1 = vec_splat_u16(1);
v8 = vec_splat_u16(8);
998
999
/* set the alpha to 255 on the destination surf */
1000
valphamask = VEC_ALPHA_MASK();