aboutsummaryrefslogtreecommitdiffhomepage
path: root/src/opts/SkSwizzler_opts.h
blob: 1d3cc51c3764c360788243fb19f3b9dd9d0c36d0 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
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
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
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
317
318
319
320
321
322
323
324
325
326
327
328
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
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
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
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
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
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
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
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
/*
 * Copyright 2016 Google Inc.
 *
 * Use of this source code is governed by a BSD-style license that can be
 * found in the LICENSE file.
 */

#ifndef SkSwizzler_opts_DEFINED
#define SkSwizzler_opts_DEFINED

#include "SkColorPriv.h"

namespace SK_OPTS_NS {

static void RGBA_to_rgbA_portable(uint32_t* dst, const void* vsrc, int count) {
    auto src = (const uint32_t*)vsrc;
    for (int i = 0; i < count; i++) {
        uint8_t a = src[i] >> 24,
                b = src[i] >> 16,
                g = src[i] >>  8,
                r = src[i] >>  0;
        b = (b*a+127)/255;
        g = (g*a+127)/255;
        r = (r*a+127)/255;
        dst[i] = (uint32_t)a << 24
               | (uint32_t)b << 16
               | (uint32_t)g <<  8
               | (uint32_t)r <<  0;
    }
}

static void RGBA_to_bgrA_portable(uint32_t* dst, const void* vsrc, int count) {
    auto src = (const uint32_t*)vsrc;
    for (int i = 0; i < count; i++) {
        uint8_t a = src[i] >> 24,
                b = src[i] >> 16,
                g = src[i] >>  8,
                r = src[i] >>  0;
        b = (b*a+127)/255;
        g = (g*a+127)/255;
        r = (r*a+127)/255;
        dst[i] = (uint32_t)a << 24
               | (uint32_t)r << 16
               | (uint32_t)g <<  8
               | (uint32_t)b <<  0;
    }
}

static void RGBA_to_BGRA_portable(uint32_t* dst, const void* vsrc, int count) {
    auto src = (const uint32_t*)vsrc;
    for (int i = 0; i < count; i++) {
        uint8_t a = src[i] >> 24,
                b = src[i] >> 16,
                g = src[i] >>  8,
                r = src[i] >>  0;
        dst[i] = (uint32_t)a << 24
               | (uint32_t)r << 16
               | (uint32_t)g <<  8
               | (uint32_t)b <<  0;
    }
}

static void RGB_to_RGB1_portable(uint32_t dst[], const void* vsrc, int count) {
    const uint8_t* src = (const uint8_t*)vsrc;
    for (int i = 0; i < count; i++) {
        uint8_t r = src[0],
                g = src[1],
                b = src[2];
        src += 3;
        dst[i] = (uint32_t)0xFF << 24
               | (uint32_t)b    << 16
               | (uint32_t)g    <<  8
               | (uint32_t)r    <<  0;
    }
}

static void RGB_to_BGR1_portable(uint32_t dst[], const void* vsrc, int count) {
    const uint8_t* src = (const uint8_t*)vsrc;
    for (int i = 0; i < count; i++) {
        uint8_t r = src[0],
                g = src[1],
                b = src[2];
        src += 3;
        dst[i] = (uint32_t)0xFF << 24
               | (uint32_t)r    << 16
               | (uint32_t)g    <<  8
               | (uint32_t)b    <<  0;
    }
}

static void gray_to_RGB1_portable(uint32_t dst[], const void* vsrc, int count) {
    const uint8_t* src = (const uint8_t*)vsrc;
    for (int i = 0; i < count; i++) {
        dst[i] = (uint32_t)0xFF   << 24
               | (uint32_t)src[i] << 16
               | (uint32_t)src[i] <<  8
               | (uint32_t)src[i] <<  0;
    }
}

static void grayA_to_RGBA_portable(uint32_t dst[], const void* vsrc, int count) {
    const uint8_t* src = (const uint8_t*)vsrc;
    for (int i = 0; i < count; i++) {
        uint8_t g = src[0],
                a = src[1];
        src += 2;
        dst[i] = (uint32_t)a << 24
               | (uint32_t)g << 16
               | (uint32_t)g <<  8
               | (uint32_t)g <<  0;
    }
}

static void grayA_to_rgbA_portable(uint32_t dst[], const void* vsrc, int count) {
    const uint8_t* src = (const uint8_t*)vsrc;
    for (int i = 0; i < count; i++) {
        uint8_t g = src[0],
                a = src[1];
        src += 2;
        g = (g*a+127)/255;
        dst[i] = (uint32_t)a << 24
               | (uint32_t)g << 16
               | (uint32_t)g <<  8
               | (uint32_t)g <<  0;
    }
}

#if defined(SK_ARM_HAS_NEON)

// Rounded divide by 255, (x + 127) / 255
static uint8x8_t div255_round(uint16x8_t x) {
    // result = (x + 127) / 255
    // result = (x + 127) / 256 + error1
    //
    // error1 = (x + 127) / (255 * 256)
    // error1 = (x + 127) / (256 * 256) + error2
    //
    // error2 = (x + 127) / (255 * 256 * 256)
    //
    // The maximum value of error2 is too small to matter.  Thus:
    // result = (x + 127) / 256 + (x + 127) / (256 * 256)
    // result = ((x + 127) / 256 + x + 127) / 256
    // result = ((x + 127) >> 8 + x + 127) >> 8
    //
    // Use >>> to represent "rounded right shift" which, conveniently,
    // NEON supports in one instruction.
    // result = ((x >>> 8) + x) >>> 8
    //
    // Note that the second right shift is actually performed as an
    // "add, round, and narrow back to 8-bits" instruction.
    return vraddhn_u16(x, vrshrq_n_u16(x, 8));
}

// Scale a byte by another, (x * y + 127) / 255
static uint8x8_t scale(uint8x8_t x, uint8x8_t y) {
    return div255_round(vmull_u8(x, y));
}

template <bool kSwapRB>
static void premul_should_swapRB(uint32_t* dst, const void* vsrc, int count) {
    auto src = (const uint32_t*)vsrc;
    while (count >= 8) {
        // Load 8 pixels.
        uint8x8x4_t rgba = vld4_u8((const uint8_t*) src);

        uint8x8_t a = rgba.val[3],
                  b = rgba.val[2],
                  g = rgba.val[1],
                  r = rgba.val[0];

        // Premultiply.
        b = scale(b, a);
        g = scale(g, a);
        r = scale(r, a);

        // Store 8 premultiplied pixels.
        if (kSwapRB) {
            rgba.val[2] = r;
            rgba.val[1] = g;
            rgba.val[0] = b;
        } else {
            rgba.val[2] = b;
            rgba.val[1] = g;
            rgba.val[0] = r;
        }
        vst4_u8((uint8_t*) dst, rgba);
        src += 8;
        dst += 8;
        count -= 8;
    }

    // Call portable code to finish up the tail of [0,8) pixels.
    auto proc = kSwapRB ? RGBA_to_bgrA_portable : RGBA_to_rgbA_portable;
    proc(dst, src, count);
}

static void RGBA_to_rgbA(uint32_t* dst, const void* src, int count) {
    premul_should_swapRB<false>(dst, src, count);
}

static void RGBA_to_bgrA(uint32_t* dst, const void* src, int count) {
    premul_should_swapRB<true>(dst, src, count);
}

static void RGBA_to_BGRA(uint32_t* dst, const void* vsrc, int count) {
    auto src = (const uint32_t*)vsrc;
    while (count >= 16) {
        // Load 16 pixels.
        uint8x16x4_t rgba = vld4q_u8((const uint8_t*) src);

        // Swap r and b.
        SkTSwap(rgba.val[0], rgba.val[2]);

        // Store 16 pixels.
        vst4q_u8((uint8_t*) dst, rgba);
        src += 16;
        dst += 16;
        count -= 16;
    }

    if (count >= 8) {
        // Load 8 pixels.
        uint8x8x4_t rgba = vld4_u8((const uint8_t*) src);

        // Swap r and b.
        SkTSwap(rgba.val[0], rgba.val[2]);

        // Store 8 pixels.
        vst4_u8((uint8_t*) dst, rgba);
        src += 8;
        dst += 8;
        count -= 8;
    }

    RGBA_to_BGRA_portable(dst, src, count);
}

template <bool kSwapRB>
static void insert_alpha_should_swaprb(uint32_t dst[], const void* vsrc, int count) {
    const uint8_t* src = (const uint8_t*) vsrc;
    while (count >= 16) {
        // Load 16 pixels.
        uint8x16x3_t rgb = vld3q_u8(src);

        // Insert an opaque alpha channel and swap if needed.
        uint8x16x4_t rgba;
        if (kSwapRB) {
            rgba.val[0] = rgb.val[2];
            rgba.val[2] = rgb.val[0];
        } else {
            rgba.val[0] = rgb.val[0];
            rgba.val[2] = rgb.val[2];
        }
        rgba.val[1] = rgb.val[1];
        rgba.val[3] = vdupq_n_u8(0xFF);

        // Store 16 pixels.
        vst4q_u8((uint8_t*) dst, rgba);
        src += 16*3;
        dst += 16;
        count -= 16;
    }

    if (count >= 8) {
        // Load 8 pixels.
        uint8x8x3_t rgb = vld3_u8(src);

        // Insert an opaque alpha channel and swap if needed.
        uint8x8x4_t rgba;
        if (kSwapRB) {
            rgba.val[0] = rgb.val[2];
            rgba.val[2] = rgb.val[0];
        } else {
            rgba.val[0] = rgb.val[0];
            rgba.val[2] = rgb.val[2];
        }
        rgba.val[1] = rgb.val[1];
        rgba.val[3] = vdup_n_u8(0xFF);

        // Store 8 pixels.
        vst4_u8((uint8_t*) dst, rgba);
        src += 8*3;
        dst += 8;
        count -= 8;
    }

    // Call portable code to finish up the tail of [0,8) pixels.
    auto proc = kSwapRB ? RGB_to_BGR1_portable : RGB_to_RGB1_portable;
    proc(dst, src, count);
}

static void RGB_to_RGB1(uint32_t dst[], const void* src, int count) {
    insert_alpha_should_swaprb<false>(dst, src, count);
}

static void RGB_to_BGR1(uint32_t dst[], const void* src, int count) {
    insert_alpha_should_swaprb<true>(dst, src, count);
}

static void gray_to_RGB1(uint32_t dst[], const void* vsrc, int count) {
    const uint8_t* src = (const uint8_t*) vsrc;
    while (count >= 16) {
        // Load 16 pixels.
        uint8x16_t gray = vld1q_u8(src);

        // Set each of the color channels.
        uint8x16x4_t rgba;
        rgba.val[0] = gray;
        rgba.val[1] = gray;
        rgba.val[2] = gray;
        rgba.val[3] = vdupq_n_u8(0xFF);

        // Store 16 pixels.
        vst4q_u8((uint8_t*) dst, rgba);
        src += 16;
        dst += 16;
        count -= 16;
    }

    if (count >= 8) {
        // Load 8 pixels.
        uint8x8_t gray = vld1_u8(src);

        // Set each of the color channels.
        uint8x8x4_t rgba;
        rgba.val[0] = gray;
        rgba.val[1] = gray;
        rgba.val[2] = gray;
        rgba.val[3] = vdup_n_u8(0xFF);

        // Store 8 pixels.
        vst4_u8((uint8_t*) dst, rgba);
        src += 8;
        dst += 8;
        count -= 8;
    }

    gray_to_RGB1_portable(dst, src, count);
}

template <bool kPremul>
static void expand_grayA(uint32_t dst[], const void* vsrc, int count) {
    const uint8_t* src = (const uint8_t*) vsrc;
    while (count >= 16) {
        // Load 16 pixels.
        uint8x16x2_t ga = vld2q_u8(src);

        // Premultiply if requested.
        if (kPremul) {
            ga.val[0] = vcombine_u8(
                    scale(vget_low_u8(ga.val[0]),  vget_low_u8(ga.val[1])),
                    scale(vget_high_u8(ga.val[0]), vget_high_u8(ga.val[1])));
        }

        // Set each of the color channels.
        uint8x16x4_t rgba;
        rgba.val[0] = ga.val[0];
        rgba.val[1] = ga.val[0];
        rgba.val[2] = ga.val[0];
        rgba.val[3] = ga.val[1];

        // Store 16 pixels.
        vst4q_u8((uint8_t*) dst, rgba);
        src += 16*2;
        dst += 16;
        count -= 16;
    }

    if (count >= 8) {
        // Load 8 pixels.
        uint8x8x2_t ga = vld2_u8(src);

        // Premultiply if requested.
        if (kPremul) {
            ga.val[0] = scale(ga.val[0], ga.val[1]);
        }

        // Set each of the color channels.
        uint8x8x4_t rgba;
        rgba.val[0] = ga.val[0];
        rgba.val[1] = ga.val[0];
        rgba.val[2] = ga.val[0];
        rgba.val[3] = ga.val[1];

        // Store 8 pixels.
        vst4_u8((uint8_t*) dst, rgba);
        src += 8*2;
        dst += 8;
        count -= 8;
    }

    auto proc = kPremul ? grayA_to_rgbA_portable : grayA_to_RGBA_portable;
    proc(dst, src, count);
}

static void grayA_to_RGBA(uint32_t dst[], const void* src, int count) {
    expand_grayA<false>(dst, src, count);
}

static void grayA_to_rgbA(uint32_t dst[], const void* src, int count) {
    expand_grayA<true>(dst, src, count);
}

#elif SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSSE3

// Scale a byte by another.
// Inputs are stored in 16-bit lanes, but are not larger than 8-bits.
static __m128i scale(__m128i x, __m128i y) {
    const __m128i _128 = _mm_set1_epi16(128);
    const __m128i _257 = _mm_set1_epi16(257);

    // (x+127)/255 == ((x+128)*257)>>16 for 0 <= x <= 255*255.
    return _mm_mulhi_epu16(_mm_add_epi16(_mm_mullo_epi16(x, y), _128), _257);
}

template <bool kSwapRB>
static void premul_should_swapRB(uint32_t* dst, const void* vsrc, int count) {
    auto src = (const uint32_t*)vsrc;

    auto premul8 = [](__m128i* lo, __m128i* hi) {
        const __m128i zeros = _mm_setzero_si128();
        __m128i planar;
        if (kSwapRB) {
            planar = _mm_setr_epi8(2,6,10,14, 1,5,9,13, 0,4,8,12, 3,7,11,15);
        } else {
            planar = _mm_setr_epi8(0,4,8,12, 1,5,9,13, 2,6,10,14, 3,7,11,15);
        }

        // Swizzle the pixels to 8-bit planar.
        *lo = _mm_shuffle_epi8(*lo, planar);                      // rrrrgggg bbbbaaaa
        *hi = _mm_shuffle_epi8(*hi, planar);                      // RRRRGGGG BBBBAAAA
        __m128i rg = _mm_unpacklo_epi32(*lo, *hi),                // rrrrRRRR ggggGGGG
                ba = _mm_unpackhi_epi32(*lo, *hi);                // bbbbBBBB aaaaAAAA

        // Unpack to 16-bit planar.
        __m128i r = _mm_unpacklo_epi8(rg, zeros),                 // r_r_r_r_ R_R_R_R_
                g = _mm_unpackhi_epi8(rg, zeros),                 // g_g_g_g_ G_G_G_G_
                b = _mm_unpacklo_epi8(ba, zeros),                 // b_b_b_b_ B_B_B_B_
                a = _mm_unpackhi_epi8(ba, zeros);                 // a_a_a_a_ A_A_A_A_

        // Premultiply!
        r = scale(r, a);
        g = scale(g, a);
        b = scale(b, a);

        // Repack into interlaced pixels.
        rg = _mm_or_si128(r, _mm_slli_epi16(g, 8));               // rgrgrgrg RGRGRGRG
        ba = _mm_or_si128(b, _mm_slli_epi16(a, 8));               // babababa BABABABA
        *lo = _mm_unpacklo_epi16(rg, ba);                         // rgbargba rgbargba
        *hi = _mm_unpackhi_epi16(rg, ba);                         // RGBARGBA RGBARGBA
    };

    while (count >= 8) {
        __m128i lo = _mm_loadu_si128((const __m128i*) (src + 0)),
                hi = _mm_loadu_si128((const __m128i*) (src + 4));

        premul8(&lo, &hi);

        _mm_storeu_si128((__m128i*) (dst + 0), lo);
        _mm_storeu_si128((__m128i*) (dst + 4), hi);

        src += 8;
        dst += 8;
        count -= 8;
    }

    if (count >= 4) {
        __m128i lo = _mm_loadu_si128((const __m128i*) src),
                hi = _mm_setzero_si128();

        premul8(&lo, &hi);

        _mm_storeu_si128((__m128i*) dst, lo);

        src += 4;
        dst += 4;
        count -= 4;
    }

    // Call portable code to finish up the tail of [0,4) pixels.
    auto proc = kSwapRB ? RGBA_to_bgrA_portable : RGBA_to_rgbA_portable;
    proc(dst, src, count);
}

static void RGBA_to_rgbA(uint32_t* dst, const void* src, int count) {
    premul_should_swapRB<false>(dst, src, count);
}

static void RGBA_to_bgrA(uint32_t* dst, const void* src, int count) {
    premul_should_swapRB<true>(dst, src, count);
}

static void RGBA_to_BGRA(uint32_t* dst, const void* vsrc, int count) {
    auto src = (const uint32_t*)vsrc;
    const __m128i swapRB = _mm_setr_epi8(2,1,0,3, 6,5,4,7, 10,9,8,11, 14,13,12,15);

    while (count >= 4) {
        __m128i rgba = _mm_loadu_si128((const __m128i*) src);
        __m128i bgra = _mm_shuffle_epi8(rgba, swapRB);
        _mm_storeu_si128((__m128i*) dst, bgra);

        src += 4;
        dst += 4;
        count -= 4;
    }

    RGBA_to_BGRA_portable(dst, src, count);
}

template <bool kSwapRB>
static void insert_alpha_should_swaprb(uint32_t dst[], const void* vsrc, int count) {
    const uint8_t* src = (const uint8_t*) vsrc;

    const __m128i alphaMask = _mm_set1_epi32(0xFF000000);
    __m128i expand;
    const uint8_t X = 0xFF; // Used a placeholder.  The value of X is irrelevant.
    if (kSwapRB) {
        expand = _mm_setr_epi8(2,1,0,X, 5,4,3,X, 8,7,6,X, 11,10,9,X);
    } else {
        expand = _mm_setr_epi8(0,1,2,X, 3,4,5,X, 6,7,8,X, 9,10,11,X);
    }

    while (count >= 6) {
        // Load a vector.  While this actually contains 5 pixels plus an
        // extra component, we will discard all but the first four pixels on
        // this iteration.
        __m128i rgb = _mm_loadu_si128((const __m128i*) src);

        // Expand the first four pixels to RGBX and then mask to RGB(FF).
        __m128i rgba = _mm_or_si128(_mm_shuffle_epi8(rgb, expand), alphaMask);

        // Store 4 pixels.
        _mm_storeu_si128((__m128i*) dst, rgba);

        src += 4*3;
        dst += 4;
        count -= 4;
    }

    // Call portable code to finish up the tail of [0,4) pixels.
    auto proc = kSwapRB ? RGB_to_BGR1_portable : RGB_to_RGB1_portable;
    proc(dst, src, count);
}

static void RGB_to_RGB1(uint32_t dst[], const void* src, int count) {
    insert_alpha_should_swaprb<false>(dst, src, count);
}

static void RGB_to_BGR1(uint32_t dst[], const void* src, int count) {
    insert_alpha_should_swaprb<true>(dst, src, count);
}

static void gray_to_RGB1(uint32_t dst[], const void* vsrc, int count) {
    const uint8_t* src = (const uint8_t*) vsrc;

    const __m128i alphas = _mm_set1_epi8((uint8_t) 0xFF);
    while (count >= 16) {
        __m128i grays = _mm_loadu_si128((const __m128i*) src);

        __m128i gg_lo = _mm_unpacklo_epi8(grays, grays);
        __m128i gg_hi = _mm_unpackhi_epi8(grays, grays);
        __m128i ga_lo = _mm_unpacklo_epi8(grays, alphas);
        __m128i ga_hi = _mm_unpackhi_epi8(grays, alphas);

        __m128i ggga0 = _mm_unpacklo_epi16(gg_lo, ga_lo);
        __m128i ggga1 = _mm_unpackhi_epi16(gg_lo, ga_lo);
        __m128i ggga2 = _mm_unpacklo_epi16(gg_hi, ga_hi);
        __m128i ggga3 = _mm_unpackhi_epi16(gg_hi, ga_hi);

        _mm_storeu_si128((__m128i*) (dst +  0), ggga0);
        _mm_storeu_si128((__m128i*) (dst +  4), ggga1);
        _mm_storeu_si128((__m128i*) (dst +  8), ggga2);
        _mm_storeu_si128((__m128i*) (dst + 12), ggga3);

        src += 16;
        dst += 16;
        count -= 16;
    }

    gray_to_RGB1_portable(dst, src, count);
}

static void grayA_to_RGBA(uint32_t dst[], const void* vsrc, int count) {
    const uint8_t* src = (const uint8_t*) vsrc;
    while (count >= 8) {
        __m128i ga = _mm_loadu_si128((const __m128i*) src);

        __m128i gg = _mm_or_si128(_mm_and_si128(ga, _mm_set1_epi16(0x00FF)),
                                  _mm_slli_epi16(ga, 8));

        __m128i ggga_lo = _mm_unpacklo_epi16(gg, ga);
        __m128i ggga_hi = _mm_unpackhi_epi16(gg, ga);

        _mm_storeu_si128((__m128i*) (dst +  0), ggga_lo);
        _mm_storeu_si128((__m128i*) (dst +  4), ggga_hi);

        src += 8*2;
        dst += 8;
        count -= 8;
    }

    grayA_to_RGBA_portable(dst, src, count);
}

static void grayA_to_rgbA(uint32_t dst[], const void* vsrc, int count) {
    const uint8_t* src = (const uint8_t*) vsrc;
    while (count >= 8) {
        __m128i grayA = _mm_loadu_si128((const __m128i*) src);

        __m128i g0 = _mm_and_si128(grayA, _mm_set1_epi16(0x00FF));
        __m128i a0 = _mm_srli_epi16(grayA, 8);

        // Premultiply
        g0 = scale(g0, a0);

        __m128i gg = _mm_or_si128(g0, _mm_slli_epi16(g0, 8));
        __m128i ga = _mm_or_si128(g0, _mm_slli_epi16(a0, 8));


        __m128i ggga_lo = _mm_unpacklo_epi16(gg, ga);
        __m128i ggga_hi = _mm_unpackhi_epi16(gg, ga);

        _mm_storeu_si128((__m128i*) (dst +  0), ggga_lo);
        _mm_storeu_si128((__m128i*) (dst +  4), ggga_hi);

        src += 8*2;
        dst += 8;
        count -= 8;
    }

    grayA_to_rgbA_portable(dst, src, count);
}

#else

static void RGBA_to_rgbA(uint32_t* dst, const void* src, int count) {
    RGBA_to_rgbA_portable(dst, src, count);
}

static void RGBA_to_bgrA(uint32_t* dst, const void* src, int count) {
    RGBA_to_bgrA_portable(dst, src, count);
}

static void RGBA_to_BGRA(uint32_t* dst, const void* src, int count) {
    RGBA_to_BGRA_portable(dst, src, count);
}

static void RGB_to_RGB1(uint32_t dst[], const void* src, int count) {
    RGB_to_RGB1_portable(dst, src, count);
}

static void RGB_to_BGR1(uint32_t dst[], const void* src, int count) {
    RGB_to_BGR1_portable(dst, src, count);
}

static void gray_to_RGB1(uint32_t dst[], const void* src, int count) {
    gray_to_RGB1_portable(dst, src, count);
}

static void grayA_to_RGBA(uint32_t dst[], const void* src, int count) {
    grayA_to_RGBA_portable(dst, src, count);
}

static void grayA_to_rgbA(uint32_t dst[], const void* src, int count) {
    grayA_to_rgbA_portable(dst, src, count);
}

#endif

}

#endif // SkSwizzler_opts_DEFINED