This repository has been archived by the owner on Feb 11, 2021. It is now read-only.
/
SDL_blit_A.c
1359 lines (1219 loc) · 40.6 KB
1
2
/*
SDL - Simple DirectMedia Layer
3
Copyright (C) 1997-2011 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.
14
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
27
28
29
#include "SDL_video.h"
#include "SDL_blit.h"
/* Functions to perform alpha blended blitting */
/* N->1 blending with per-surface alpha */
30
31
static void
BlitNto1SurfaceAlpha(SDL_BlitInfo * info)
32
{
33
34
35
int width = info->dst_w;
int height = info->dst_h;
Uint8 *src = info->src;
36
int srcskip = info->src_skip;
37
Uint8 *dst = info->dst;
38
int dstskip = info->dst_skip;
39
Uint8 *palmap = info->table;
40
41
SDL_PixelFormat *srcfmt = info->src_fmt;
SDL_PixelFormat *dstfmt = info->dst_fmt;
42
43
int srcbpp = srcfmt->BytesPerPixel;
44
const unsigned A = info->a;
45
46
47
while (height--) {
/* *INDENT-OFF* */
48
49
DUFFS_LOOP4(
{
50
Uint32 Pixel;
51
52
53
54
55
56
unsigned sR;
unsigned sG;
unsigned sB;
unsigned dR;
unsigned dG;
unsigned dB;
57
DISEMBLE_RGB(src, srcbpp, srcfmt, Pixel, sR, sG, sB);
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
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);
79
80
81
82
/* *INDENT-ON* */
src += srcskip;
dst += dstskip;
}
83
84
85
}
/* N->1 blending with pixel alpha */
86
87
static void
BlitNto1PixelAlpha(SDL_BlitInfo * info)
88
{
89
90
91
int width = info->dst_w;
int height = info->dst_h;
Uint8 *src = info->src;
92
int srcskip = info->src_skip;
93
Uint8 *dst = info->dst;
94
int dstskip = info->dst_skip;
95
Uint8 *palmap = info->table;
96
97
SDL_PixelFormat *srcfmt = info->src_fmt;
SDL_PixelFormat *dstfmt = info->dst_fmt;
98
99
100
101
102
int srcbpp = srcfmt->BytesPerPixel;
/* FIXME: fix alpha bit field expansion here too? */
while (height--) {
/* *INDENT-OFF* */
103
104
DUFFS_LOOP4(
{
105
Uint32 Pixel;
106
107
108
109
110
111
112
unsigned sR;
unsigned sG;
unsigned sB;
unsigned sA;
unsigned dR;
unsigned dG;
unsigned dB;
113
DISEMBLE_RGBA(src,srcbpp,srcfmt,Pixel,sR,sG,sB,sA);
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
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);
135
136
137
138
/* *INDENT-ON* */
src += srcskip;
dst += dstskip;
}
139
140
141
}
/* colorkeyed N->1 blending with per-surface alpha */
142
143
static void
BlitNto1SurfaceAlphaKey(SDL_BlitInfo * info)
144
{
145
146
147
int width = info->dst_w;
int height = info->dst_h;
Uint8 *src = info->src;
148
int srcskip = info->src_skip;
149
Uint8 *dst = info->dst;
150
int dstskip = info->dst_skip;
151
Uint8 *palmap = info->table;
152
153
SDL_PixelFormat *srcfmt = info->src_fmt;
SDL_PixelFormat *dstfmt = info->dst_fmt;
154
int srcbpp = srcfmt->BytesPerPixel;
155
Uint32 ckey = info->colorkey;
156
157
const int A = info->a;
158
159
160
while (height--) {
/* *INDENT-OFF* */
161
162
DUFFS_LOOP(
{
163
Uint32 Pixel;
164
165
166
167
168
169
unsigned sR;
unsigned sG;
unsigned sB;
unsigned dR;
unsigned dG;
unsigned dB;
170
171
DISEMBLE_RGB(src, srcbpp, srcfmt, Pixel, sR, sG, sB);
if ( Pixel != ckey ) {
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
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);
194
195
196
197
/* *INDENT-ON* */
src += srcskip;
dst += dstskip;
}
198
199
}
200
#ifdef __MMX__
201
202
/* fast RGB888->(A)RGB888 blending with surface alpha=128 special case */
203
204
static void
BlitRGBtoRGBSurfaceAlpha128MMX(SDL_BlitInfo * info)
205
{
206
207
208
int width = info->dst_w;
int height = info->dst_h;
Uint32 *srcp = (Uint32 *) info->src;
209
int srcskip = info->src_skip >> 2;
210
Uint32 *dstp = (Uint32 *) info->dst;
211
212
int dstskip = info->dst_skip >> 2;
Uint32 dalpha = info->dst_fmt->Amask;
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
__m64 src1, src2, dst1, dst2, lmask, hmask, dsta;
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 */
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) */
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 */
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 */
*(__m64 *) dstp = dst1; /* dst1 -> 2 x dst pixels */
dstp += 2;
srcp += 2;
}
srcp += srcskip;
dstp += dstskip;
}
_mm_empty();
256
257
258
}
/* fast RGB888->(A)RGB888 blending with surface alpha */
259
260
static void
BlitRGBtoRGBSurfaceAlphaMMX(SDL_BlitInfo * info)
261
{
262
SDL_PixelFormat *df = info->dst_fmt;
263
Uint32 chanmask = df->Rmask | df->Gmask | df->Bmask;
264
unsigned alpha = info->a;
265
266
267
268
269
if (alpha == 128 && (df->Rmask | df->Gmask | df->Bmask) == 0x00FFFFFF) {
/* only call a128 version when R,G,B occupy lower bits */
BlitRGBtoRGBSurfaceAlpha128MMX(info);
} else {
270
271
272
int width = info->dst_w;
int height = info->dst_h;
Uint32 *srcp = (Uint32 *) info->src;
273
int srcskip = info->src_skip >> 2;
274
Uint32 *dstp = (Uint32 *) info->dst;
275
int dstskip = info->dst_skip >> 2;
276
277
278
279
280
281
282
283
284
285
Uint32 dalpha = df->Amask;
Uint32 amult;
__m64 src1, src2, dst1, dst2, mm_alpha, mm_zero, dsta;
mm_zero = _mm_setzero_si64(); /* 0 -> mm_zero */
/* form the alpha mult */
amult = alpha | (alpha << 8);
amult = amult | (amult << 16);
chanmask =
286
287
(0xff << df->Rshift) | (0xff << df->
Gshift) | (0xff << df->Bshift);
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
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 */
/* at this point mm_alpha can be 000A0A0A or 0A0A0A00 or another combo */
dsta = _mm_set_pi32(dalpha, dalpha); /* dst alpha mask -> dsta */
while (height--) {
int n = width;
if (n & 1) {
/* One Pixel Blend */
src2 = _mm_cvtsi32_si64(*srcp); /* src(ARGB) -> src2 (0000ARGB) */
src2 = _mm_unpacklo_pi8(src2, mm_zero); /* 0A0R0G0B -> src2 */
dst1 = _mm_cvtsi32_si64(*dstp); /* dst(ARGB) -> dst1 (0000ARGB) */
dst1 = _mm_unpacklo_pi8(dst1, mm_zero); /* 0A0R0G0B -> dst1 */
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 */
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 */
++srcp;
++dstp;
n--;
}
317
318
319
320
321
322
323
for (n >>= 1; n > 0; --n) {
/* Two Pixels Blend */
src1 = *(__m64 *) srcp; /* 2 x src -> src1(ARGBARGB) */
src2 = src1; /* 2 x src -> src2(ARGBARGB) */
src1 = _mm_unpacklo_pi8(src1, mm_zero); /* low - 0A0R0G0B -> src1 */
src2 = _mm_unpackhi_pi8(src2, mm_zero); /* high - 0A0R0G0B -> src2 */
324
325
326
327
328
dst1 = *(__m64 *) dstp; /* 2 x dst -> dst1(ARGBARGB) */
dst2 = dst1; /* 2 x dst -> dst2(ARGBARGB) */
dst1 = _mm_unpacklo_pi8(dst1, mm_zero); /* low - 0A0R0G0B -> dst1 */
dst2 = _mm_unpackhi_pi8(dst2, mm_zero); /* high - 0A0R0G0B -> dst2 */
329
330
331
332
333
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 */
334
335
336
337
338
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 */
339
340
341
342
343
344
345
346
347
348
349
350
351
352
dst1 = _mm_packs_pu16(dst1, dst2); /* 0A0R0G0B(res1), 0A0R0G0B(res2) -> dst1(ARGBARGB) */
dst1 = _mm_or_si64(dst1, dsta); /* dsta | dst1 -> dst1 */
*(__m64 *) dstp = dst1; /* dst1 -> 2 x pixel */
srcp += 2;
dstp += 2;
}
srcp += srcskip;
dstp += dstskip;
}
_mm_empty();
}
353
354
355
}
/* fast ARGB888->(A)RGB888 blending with pixel alpha */
356
357
static void
BlitRGBtoRGBPixelAlphaMMX(SDL_BlitInfo * info)
358
{
359
360
361
int width = info->dst_w;
int height = info->dst_h;
Uint32 *srcp = (Uint32 *) info->src;
362
int srcskip = info->src_skip >> 2;
363
Uint32 *dstp = (Uint32 *) info->dst;
364
365
int dstskip = info->dst_skip >> 2;
SDL_PixelFormat *sf = info->src_fmt;
366
367
368
369
370
371
372
373
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;
mm_zero = _mm_setzero_si64(); /* 0 -> mm_zero */
374
multmask = 0xFFFF;
375
376
multmask <<= (ashift * 2);
multmask = ~multmask;
377
378
379
380
dmask = *(__m64 *) & multmask; /* dst alpha mask -> dmask */
while (height--) {
/* *INDENT-OFF* */
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
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);
413
414
415
416
417
/* *INDENT-ON* */
srcp += srcskip;
dstp += dstskip;
}
_mm_empty();
418
}
419
420
#endif /* __MMX__ */
421
422
/* fast RGB888->(A)RGB888 blending with surface alpha=128 special case */
423
424
static void
BlitRGBtoRGBSurfaceAlpha128(SDL_BlitInfo * info)
425
{
426
427
428
int width = info->dst_w;
int height = info->dst_h;
Uint32 *srcp = (Uint32 *) info->src;
429
int srcskip = info->src_skip >> 2;
430
Uint32 *dstp = (Uint32 *) info->dst;
431
int dstskip = info->dst_skip >> 2;
432
433
434
while (height--) {
/* *INDENT-OFF* */
435
DUFFS_LOOP4({
436
437
438
439
Uint32 s = *srcp++;
Uint32 d = *dstp;
*dstp++ = ((((s & 0x00fefefe) + (d & 0x00fefefe)) >> 1)
+ (s & d & 0x00010101)) | 0xff000000;
440
}, width);
441
442
443
444
/* *INDENT-ON* */
srcp += srcskip;
dstp += dstskip;
}
445
446
}
447
/* fast RGB888->(A)RGB888 blending with surface alpha */
448
449
static void
BlitRGBtoRGBSurfaceAlpha(SDL_BlitInfo * info)
450
{
451
unsigned alpha = info->a;
452
453
454
if (alpha == 128) {
BlitRGBtoRGBSurfaceAlpha128(info);
} else {
455
456
457
int width = info->dst_w;
int height = info->dst_h;
Uint32 *srcp = (Uint32 *) info->src;
458
int srcskip = info->src_skip >> 2;
459
Uint32 *dstp = (Uint32 *) info->dst;
460
int dstskip = info->dst_skip >> 2;
461
462
463
464
465
466
467
Uint32 s;
Uint32 d;
Uint32 s1;
Uint32 d1;
while (height--) {
/* *INDENT-OFF* */
468
DUFFS_LOOP4({
469
470
471
472
473
474
475
476
477
478
479
480
481
s = *srcp;
d = *dstp;
s1 = s & 0xff00ff;
d1 = d & 0xff00ff;
d1 = (d1 + ((s1 - d1) * alpha >> 8))
& 0xff00ff;
s &= 0xff00;
d &= 0xff00;
d = (d + ((s - d) * alpha >> 8)) & 0xff00;
*dstp = d1 | d | 0xff000000;
++srcp;
++dstp;
}, width);
482
483
484
485
486
/* *INDENT-ON* */
srcp += srcskip;
dstp += dstskip;
}
}
487
488
}
489
/* fast ARGB888->(A)RGB888 blending with pixel alpha */
490
491
static void
BlitRGBtoRGBPixelAlpha(SDL_BlitInfo * info)
492
{
493
494
495
int width = info->dst_w;
int height = info->dst_h;
Uint32 *srcp = (Uint32 *) info->src;
496
int srcskip = info->src_skip >> 2;
497
Uint32 *dstp = (Uint32 *) info->dst;
498
int dstskip = info->dst_skip >> 2;
499
500
501
while (height--) {
/* *INDENT-OFF* */
502
503
504
505
506
507
508
509
510
511
512
DUFFS_LOOP4({
Uint32 dalpha;
Uint32 d;
Uint32 s1;
Uint32 d1;
Uint32 s = *srcp;
Uint32 alpha = s >> 24;
/* FIXME: Here we special-case opaque alpha since the
compositioning used (>>8 instead of /255) doesn't handle
it correctly. Also special-case alpha=0 for speed?
Benchmark this! */
513
514
if(alpha) {
if(alpha == SDL_ALPHA_OPAQUE) {
515
*dstp = (s & 0x00ffffff) | (*dstp & 0xff000000);
516
} else {
517
518
519
520
521
522
523
524
525
526
527
528
529
/*
* take out the middle component (green), and process
* the other two in parallel. One multiply less.
*/
d = *dstp;
dalpha = d & 0xff000000;
s1 = s & 0xff00ff;
d1 = d & 0xff00ff;
d1 = (d1 + ((s1 - d1) * alpha >> 8)) & 0xff00ff;
s &= 0xff00;
d &= 0xff00;
d = (d + ((s - d) * alpha >> 8)) & 0xff00;
*dstp = d1 | d | dalpha;
530
}
531
532
533
534
}
++srcp;
++dstp;
}, width);
535
536
537
538
/* *INDENT-ON* */
srcp += srcskip;
dstp += dstskip;
}
539
540
}
541
542
543
544
545
546
547
548
549
550
551
/* 16bpp special case for per-surface alpha=50%: blend 2 pixels in parallel */
/* blend a single 16 bit pixel at 50% */
#define BLEND16_50(d, s, mask) \
((((s & mask) + (d & mask)) >> 1) + (s & d & (~mask & 0xffff)))
/* blend two 16 bit pixels at 50% */
#define BLEND2x16_50(d, s, mask) \
(((s & (mask | mask << 16)) >> 1) + ((d & (mask | mask << 16)) >> 1) \
+ (s & d & (~(mask | mask << 16))))
552
553
static void
Blit16to16SurfaceAlpha128(SDL_BlitInfo * info, Uint16 mask)
554
{
555
556
557
int width = info->dst_w;
int height = info->dst_h;
Uint16 *srcp = (Uint16 *) info->src;
558
int srcskip = info->src_skip >> 1;
559
Uint16 *dstp = (Uint16 *) info->dst;
560
int dstskip = info->dst_skip >> 1;
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
while (height--) {
if (((uintptr_t) srcp ^ (uintptr_t) dstp) & 2) {
/*
* Source and destination not aligned, pipeline it.
* This is mostly a win for big blits but no loss for
* small ones
*/
Uint32 prev_sw;
int w = width;
/* handle odd destination */
if ((uintptr_t) dstp & 2) {
Uint16 d = *dstp, s = *srcp;
*dstp = BLEND16_50(d, s, mask);
dstp++;
srcp++;
w--;
}
srcp++; /* srcp is now 32-bit aligned */
/* bootstrap pipeline with first halfword */
prev_sw = ((Uint32 *) srcp)[-1];
while (w > 1) {
Uint32 sw, dw, s;
sw = *(Uint32 *) srcp;
dw = *(Uint32 *) dstp;
589
#if SDL_BYTEORDER == SDL_BIG_ENDIAN
590
s = (prev_sw << 16) + (sw >> 16);
591
#else
592
s = (prev_sw >> 16) + (sw << 16);
593
#endif
594
595
596
597
598
599
600
601
602
603
prev_sw = sw;
*(Uint32 *) dstp = BLEND2x16_50(dw, s, mask);
dstp += 2;
srcp += 2;
w -= 2;
}
/* final pixel if any */
if (w) {
Uint16 d = *dstp, s;
604
#if SDL_BYTEORDER == SDL_BIG_ENDIAN
605
s = (Uint16) prev_sw;
606
#else
607
s = (Uint16) (prev_sw >> 16);
608
#endif
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
*dstp = BLEND16_50(d, s, mask);
srcp++;
dstp++;
}
srcp += srcskip - 1;
dstp += dstskip;
} else {
/* source and destination are aligned */
int w = width;
/* first odd pixel? */
if ((uintptr_t) srcp & 2) {
Uint16 d = *dstp, s = *srcp;
*dstp = BLEND16_50(d, s, mask);
srcp++;
dstp++;
w--;
}
/* srcp and dstp are now 32-bit aligned */
while (w > 1) {
Uint32 sw = *(Uint32 *) srcp;
Uint32 dw = *(Uint32 *) dstp;
*(Uint32 *) dstp = BLEND2x16_50(dw, sw, mask);
srcp += 2;
dstp += 2;
w -= 2;
}
/* last odd pixel? */
if (w) {
Uint16 d = *dstp, s = *srcp;
*dstp = BLEND16_50(d, s, mask);
srcp++;
dstp++;
}
srcp += srcskip;
dstp += dstskip;
}
}
649
650
}
651
#ifdef __MMX__
652
653
/* fast RGB565->RGB565 blending with surface alpha */
654
655
static void
Blit565to565SurfaceAlphaMMX(SDL_BlitInfo * info)
656
{
657
unsigned alpha = info->a;
658
659
660
if (alpha == 128) {
Blit16to16SurfaceAlpha128(info, 0xf7de);
} else {
661
662
663
int width = info->dst_w;
int height = info->dst_h;
Uint16 *srcp = (Uint16 *) info->src;
664
int srcskip = info->src_skip >> 1;
665
Uint16 *dstp = (Uint16 *) info->dst;
666
int dstskip = info->dst_skip >> 1;
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
Uint32 s, d;
__m64 src1, dst1, src2, dst2, gmask, bmask, mm_res, mm_alpha;
alpha &= ~(1 + 2 + 4); /* cut alpha to get the exact same behaviour */
mm_alpha = _mm_set_pi32(0, alpha); /* 0000000A -> mm_alpha */
alpha >>= 3; /* downscale alpha to 5 bits */
mm_alpha = _mm_unpacklo_pi16(mm_alpha, mm_alpha); /* 00000A0A -> mm_alpha */
mm_alpha = _mm_unpacklo_pi32(mm_alpha, mm_alpha); /* 0A0A0A0A -> mm_alpha */
/* position alpha to allow for mullo and mulhi on diff channels
to reduce the number of operations */
mm_alpha = _mm_slli_si64(mm_alpha, 3);
/* Setup the 565 color channel masks */
gmask = _mm_set_pi32(0x07E007E0, 0x07E007E0); /* MASKGREEN -> gmask */
bmask = _mm_set_pi32(0x001F001F, 0x001F001F); /* MASKBLUE -> bmask */
while (height--) {
/* *INDENT-OFF* */
687
DUFFS_LOOP_124(
688
689
690
691
692
693
694
695
696
697
698
699
{
s = *srcp++;
d = *dstp;
/*
* shift out the middle component (green) to
* the high 16 bits, and process all three RGB
* components at the same time.
*/
s = (s | s << 16) & 0x07e0f81f;
d = (d | d << 16) & 0x07e0f81f;
d += (s - d) * alpha >> 5;
d &= 0x07e0f81f;
700
*dstp++ = (Uint16)(d | d >> 16);
701
702
703
704
705
706
707
708
709
710
711
712
},{
s = *srcp++;
d = *dstp;
/*
* shift out the middle component (green) to
* the high 16 bits, and process all three RGB
* components at the same time.
*/
s = (s | s << 16) & 0x07e0f81f;
d = (d | d << 16) & 0x07e0f81f;
d += (s - d) * alpha >> 5;
d &= 0x07e0f81f;
713
*dstp++ = (Uint16)(d | d >> 16);
714
715
716
717
718
719
720
721
722
723
724
s = *srcp++;
d = *dstp;
/*
* shift out the middle component (green) to
* the high 16 bits, and process all three RGB
* components at the same time.
*/
s = (s | s << 16) & 0x07e0f81f;
d = (d | d << 16) & 0x07e0f81f;
d += (s - d) * alpha >> 5;
d &= 0x07e0f81f;
725
*dstp++ = (Uint16)(d | d >> 16);
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
},{
src1 = *(__m64*)srcp; /* 4 src pixels -> src1 */
dst1 = *(__m64*)dstp; /* 4 dst pixels -> dst1 */
/* red */
src2 = src1;
src2 = _mm_srli_pi16(src2, 11); /* src2 >> 11 -> src2 [000r 000r 000r 000r] */
dst2 = dst1;
dst2 = _mm_srli_pi16(dst2, 11); /* dst2 >> 11 -> dst2 [000r 000r 000r 000r] */
/* blend */
src2 = _mm_sub_pi16(src2, dst2);/* src - dst -> src2 */
src2 = _mm_mullo_pi16(src2, mm_alpha); /* src2 * alpha -> src2 */
src2 = _mm_srli_pi16(src2, 11); /* src2 >> 11 -> src2 */
dst2 = _mm_add_pi16(src2, dst2); /* src2 + dst2 -> dst2 */
dst2 = _mm_slli_pi16(dst2, 11); /* dst2 << 11 -> dst2 */
mm_res = dst2; /* RED -> mm_res */
/* green -- process the bits in place */
src2 = src1;
src2 = _mm_and_si64(src2, gmask); /* src & MASKGREEN -> src2 */
dst2 = dst1;
dst2 = _mm_and_si64(dst2, gmask); /* dst & MASKGREEN -> dst2 */
/* blend */
src2 = _mm_sub_pi16(src2, dst2);/* src - dst -> src2 */
src2 = _mm_mulhi_pi16(src2, mm_alpha); /* src2 * alpha -> src2 */
src2 = _mm_slli_pi16(src2, 5); /* src2 << 5 -> src2 */
dst2 = _mm_add_pi16(src2, dst2); /* src2 + dst2 -> dst2 */
mm_res = _mm_or_si64(mm_res, dst2); /* RED | GREEN -> mm_res */
/* blue */
src2 = src1;
src2 = _mm_and_si64(src2, bmask); /* src & MASKBLUE -> src2[000b 000b 000b 000b] */
dst2 = dst1;
dst2 = _mm_and_si64(dst2, bmask); /* dst & MASKBLUE -> dst2[000b 000b 000b 000b] */
/* blend */
src2 = _mm_sub_pi16(src2, dst2);/* src - dst -> src2 */
src2 = _mm_mullo_pi16(src2, mm_alpha); /* src2 * alpha -> src2 */
src2 = _mm_srli_pi16(src2, 11); /* src2 >> 11 -> src2 */
dst2 = _mm_add_pi16(src2, dst2); /* src2 + dst2 -> dst2 */
dst2 = _mm_and_si64(dst2, bmask); /* dst2 & MASKBLUE -> dst2 */
mm_res = _mm_or_si64(mm_res, dst2); /* RED | GREEN | BLUE -> mm_res */
*(__m64*)dstp = mm_res; /* mm_res -> 4 dst pixels */
srcp += 4;
dstp += 4;
781
782
783
784
785
786
787
}, width);
/* *INDENT-ON* */
srcp += srcskip;
dstp += dstskip;
}
_mm_empty();
}
788
789
790
}
/* fast RGB555->RGB555 blending with surface alpha */
791
792
static void
Blit555to555SurfaceAlphaMMX(SDL_BlitInfo * info)
793
{
794
unsigned alpha = info->a;
795
796
797
if (alpha == 128) {
Blit16to16SurfaceAlpha128(info, 0xfbde);
} else {
798
799
800
int width = info->dst_w;
int height = info->dst_h;
Uint16 *srcp = (Uint16 *) info->src;
801
int srcskip = info->src_skip >> 1;
802
Uint16 *dstp = (Uint16 *) info->dst;
803
int dstskip = info->dst_skip >> 1;
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
Uint32 s, d;
__m64 src1, dst1, src2, dst2, rmask, gmask, bmask, mm_res, mm_alpha;
alpha &= ~(1 + 2 + 4); /* cut alpha to get the exact same behaviour */
mm_alpha = _mm_set_pi32(0, alpha); /* 0000000A -> mm_alpha */
alpha >>= 3; /* downscale alpha to 5 bits */
mm_alpha = _mm_unpacklo_pi16(mm_alpha, mm_alpha); /* 00000A0A -> mm_alpha */
mm_alpha = _mm_unpacklo_pi32(mm_alpha, mm_alpha); /* 0A0A0A0A -> mm_alpha */
/* position alpha to allow for mullo and mulhi on diff channels
to reduce the number of operations */
mm_alpha = _mm_slli_si64(mm_alpha, 3);
/* Setup the 555 color channel masks */
rmask = _mm_set_pi32(0x7C007C00, 0x7C007C00); /* MASKRED -> rmask */
gmask = _mm_set_pi32(0x03E003E0, 0x03E003E0); /* MASKGREEN -> gmask */
bmask = _mm_set_pi32(0x001F001F, 0x001F001F); /* MASKBLUE -> bmask */
while (height--) {
/* *INDENT-OFF* */
825
DUFFS_LOOP_124(
826
827
828
829
830
831
832
833
834
835
836
837
{
s = *srcp++;
d = *dstp;
/*
* shift out the middle component (green) to
* the high 16 bits, and process all three RGB
* components at the same time.
*/
s = (s | s << 16) & 0x03e07c1f;
d = (d | d << 16) & 0x03e07c1f;
d += (s - d) * alpha >> 5;
d &= 0x03e07c1f;
838
*dstp++ = (Uint16)(d | d >> 16);
839
840
841
842
843
844
845
846
847
848
849
850
},{
s = *srcp++;
d = *dstp;
/*
* shift out the middle component (green) to
* the high 16 bits, and process all three RGB
* components at the same time.
*/
s = (s | s << 16) & 0x03e07c1f;
d = (d | d << 16) & 0x03e07c1f;
d += (s - d) * alpha >> 5;
d &= 0x03e07c1f;
851
*dstp++ = (Uint16)(d | d >> 16);
852
853
854
855
856
857
858
859
860
861
862
s = *srcp++;
d = *dstp;
/*
* shift out the middle component (green) to
* the high 16 bits, and process all three RGB
* components at the same time.
*/
s = (s | s << 16) & 0x03e07c1f;
d = (d | d << 16) & 0x03e07c1f;
d += (s - d) * alpha >> 5;
d &= 0x03e07c1f;
863
*dstp++ = (Uint16)(d | d >> 16);
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
},{
src1 = *(__m64*)srcp; /* 4 src pixels -> src1 */
dst1 = *(__m64*)dstp; /* 4 dst pixels -> dst1 */
/* red -- process the bits in place */
src2 = src1;
src2 = _mm_and_si64(src2, rmask); /* src & MASKRED -> src2 */
dst2 = dst1;
dst2 = _mm_and_si64(dst2, rmask); /* dst & MASKRED -> dst2 */
/* blend */
src2 = _mm_sub_pi16(src2, dst2);/* src - dst -> src2 */
src2 = _mm_mulhi_pi16(src2, mm_alpha); /* src2 * alpha -> src2 */
src2 = _mm_slli_pi16(src2, 5); /* src2 << 5 -> src2 */
dst2 = _mm_add_pi16(src2, dst2); /* src2 + dst2 -> dst2 */
dst2 = _mm_and_si64(dst2, rmask); /* dst2 & MASKRED -> dst2 */
mm_res = dst2; /* RED -> mm_res */
/* green -- process the bits in place */
src2 = src1;
src2 = _mm_and_si64(src2, gmask); /* src & MASKGREEN -> src2 */
dst2 = dst1;
dst2 = _mm_and_si64(dst2, gmask); /* dst & MASKGREEN -> dst2 */
/* blend */
src2 = _mm_sub_pi16(src2, dst2);/* src - dst -> src2 */
src2 = _mm_mulhi_pi16(src2, mm_alpha); /* src2 * alpha -> src2 */
src2 = _mm_slli_pi16(src2, 5); /* src2 << 5 -> src2 */
dst2 = _mm_add_pi16(src2, dst2); /* src2 + dst2 -> dst2 */
mm_res = _mm_or_si64(mm_res, dst2); /* RED | GREEN -> mm_res */
/* blue */
src2 = src1; /* src -> src2 */
src2 = _mm_and_si64(src2, bmask); /* src & MASKBLUE -> src2[000b 000b 000b 000b] */
dst2 = dst1; /* dst -> dst2 */
dst2 = _mm_and_si64(dst2, bmask); /* dst & MASKBLUE -> dst2[000b 000b 000b 000b] */
/* blend */
src2 = _mm_sub_pi16(src2, dst2);/* src - dst -> src2 */
src2 = _mm_mullo_pi16(src2, mm_alpha); /* src2 * alpha -> src2 */
src2 = _mm_srli_pi16(src2, 11); /* src2 >> 11 -> src2 */
dst2 = _mm_add_pi16(src2, dst2); /* src2 + dst2 -> dst2 */
dst2 = _mm_and_si64(dst2, bmask); /* dst2 & MASKBLUE -> dst2 */
mm_res = _mm_or_si64(mm_res, dst2); /* RED | GREEN | BLUE -> mm_res */
*(__m64*)dstp = mm_res; /* mm_res -> 4 dst pixels */
srcp += 4;
dstp += 4;
919
920
921
922
923
924
925
}, width);
/* *INDENT-ON* */
srcp += srcskip;
dstp += dstskip;
}
_mm_empty();
}
926
}
927
928
#endif /* __MMX__ */
929
930
/* fast RGB565->RGB565 blending with surface alpha */
931
932
static void
Blit565to565SurfaceAlpha(SDL_BlitInfo * info)
933
{
934
unsigned alpha = info->a;
935
936
937
if (alpha == 128) {
Blit16to16SurfaceAlpha128(info, 0xf7de);
} else {
938
939
940
int width = info->dst_w;
int height = info->dst_h;
Uint16 *srcp = (Uint16 *) info->src;
941
int srcskip = info->src_skip >> 1;
942
Uint16 *dstp = (Uint16 *) info->dst;
943
int dstskip = info->dst_skip >> 1;
944
945
946
947
alpha >>= 3; /* downscale alpha to 5 bits */
while (height--) {
/* *INDENT-OFF* */
948
949
950
951
952
953
954
955
956
957
958
959
DUFFS_LOOP4({
Uint32 s = *srcp++;
Uint32 d = *dstp;
/*
* shift out the middle component (green) to
* the high 16 bits, and process all three RGB
* components at the same time.
*/
s = (s | s << 16) & 0x07e0f81f;
d = (d | d << 16) & 0x07e0f81f;
d += (s - d) * alpha >> 5;
d &= 0x07e0f81f;
960
*dstp++ = (Uint16)(d | d >> 16);
961
}, width);
962
963
964
965
966
/* *INDENT-ON* */
srcp += srcskip;
dstp += dstskip;
}
}
967
968
969
}
/* fast RGB555->RGB555 blending with surface alpha */
970
971
static void
Blit555to555SurfaceAlpha(SDL_BlitInfo * info)
972
{
973
unsigned alpha = info->a; /* downscale alpha to 5 bits */
974
975
976
if (alpha == 128) {
Blit16to16SurfaceAlpha128(info, 0xfbde);
} else {
977
978
979
int width = info->dst_w;
int height = info->dst_h;
Uint16 *srcp = (Uint16 *) info->src;
980
int srcskip = info->src_skip >> 1;
981
Uint16 *dstp = (Uint16 *) info->dst;
982
int dstskip = info->dst_skip >> 1;
983
984
985
986
alpha >>= 3; /* downscale alpha to 5 bits */
while (height--) {
/* *INDENT-OFF* */
987
988
989
990
991
992
993
994
995
996
997
998
DUFFS_LOOP4({
Uint32 s = *srcp++;
Uint32 d = *dstp;
/*
* shift out the middle component (green) to
* the high 16 bits, and process all three RGB
* components at the same time.
*/
s = (s | s << 16) & 0x03e07c1f;
d = (d | d << 16) & 0x03e07c1f;
d += (s - d) * alpha >> 5;
d &= 0x03e07c1f;
999
*dstp++ = (Uint16)(d | d >> 16);
1000
}, width);