summaryrefslogtreecommitdiff
path: root/src/gui/painting/qdrawhelper_neon.cpp
blob: b1e80adc6b2f6c42859fdc5730e3209344f5d88a (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
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
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
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
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
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
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
/****************************************************************************
**
** Copyright (C) 2012 Digia Plc and/or its subsidiary(-ies).
** Contact: http://www.qt-project.org/legal
**
** This file is part of the QtGui module of the Qt Toolkit.
**
** $QT_BEGIN_LICENSE:LGPL$
** Commercial License Usage
** Licensees holding valid commercial Qt licenses may use this file in
** accordance with the commercial license agreement provided with the
** Software or, alternatively, in accordance with the terms contained in
** a written agreement between you and Digia.  For licensing terms and
** conditions see http://qt.digia.com/licensing.  For further information
** use the contact form at http://qt.digia.com/contact-us.
**
** GNU Lesser General Public License Usage
** Alternatively, this file may be used under the terms of the GNU Lesser
** General Public License version 2.1 as published by the Free Software
** Foundation and appearing in the file LICENSE.LGPL included in the
** packaging of this file.  Please review the following information to
** ensure the GNU Lesser General Public License version 2.1 requirements
** will be met: http://www.gnu.org/licenses/old-licenses/lgpl-2.1.html.
**
** In addition, as a special exception, Digia gives you certain additional
** rights.  These rights are described in the Digia Qt LGPL Exception
** version 1.1, included in the file LGPL_EXCEPTION.txt in this package.
**
** GNU General Public License Usage
** Alternatively, this file may be used under the terms of the GNU
** General Public License version 3.0 as published by the Free Software
** Foundation and appearing in the file LICENSE.GPL included in the
** packaging of this file.  Please review the following information to
** ensure the GNU General Public License version 3.0 requirements will be
** met: http://www.gnu.org/copyleft/gpl.html.
**
**
** $QT_END_LICENSE$
**
****************************************************************************/

#include <private/qdrawhelper_p.h>
#include <private/qblendfunctions_p.h>
#include <private/qmath_p.h>

#ifdef QT_HAVE_NEON

#include <private/qdrawhelper_neon_p.h>
#include <private/qpaintengine_raster_p.h>
#include <arm_neon.h>

QT_BEGIN_NAMESPACE

void qt_memfill32_neon(quint32 *dest, quint32 value, int count)
{
    const int epilogueSize = count % 16;
    if (count >= 16) {
        quint32 *const neonEnd = dest + count - epilogueSize;
        register uint32x4_t valueVector1 asm ("q0") = vdupq_n_u32(value);
        register uint32x4_t valueVector2 asm ("q1") = valueVector1;
        while (dest != neonEnd) {
            asm volatile (
                "vst2.32     { d0, d1, d2, d3 }, [%[DST]] !\n\t"
                "vst2.32     { d0, d1, d2, d3 }, [%[DST]] !\n\t"
                : [DST]"+r" (dest)
                : [VALUE1]"w"(valueVector1), [VALUE2]"w"(valueVector2)
                : "memory"
            );
        }
    }

    switch (epilogueSize)
    {
    case 15:     *dest++ = value;
    case 14:     *dest++ = value;
    case 13:     *dest++ = value;
    case 12:     *dest++ = value;
    case 11:     *dest++ = value;
    case 10:     *dest++ = value;
    case 9:      *dest++ = value;
    case 8:      *dest++ = value;
    case 7:      *dest++ = value;
    case 6:      *dest++ = value;
    case 5:      *dest++ = value;
    case 4:      *dest++ = value;
    case 3:      *dest++ = value;
    case 2:      *dest++ = value;
    case 1:      *dest++ = value;
    }
}

static inline uint16x8_t qvdiv_255_u16(uint16x8_t x, uint16x8_t half)
{
    // result = (x + (x >> 8) + 0x80) >> 8

    const uint16x8_t temp = vshrq_n_u16(x, 8); // x >> 8
    const uint16x8_t sum_part = vaddq_u16(x, half); // x + 0x80
    const uint16x8_t sum = vaddq_u16(temp, sum_part);

    return vshrq_n_u16(sum, 8);
}

static inline uint16x8_t qvbyte_mul_u16(uint16x8_t x, uint16x8_t alpha, uint16x8_t half)
{
    // t = qRound(x * alpha / 255.0)

    const uint16x8_t t = vmulq_u16(x, alpha); // t
    return qvdiv_255_u16(t, half);
}

static inline uint16x8_t qvinterpolate_pixel_255(uint16x8_t x, uint16x8_t a, uint16x8_t y, uint16x8_t b, uint16x8_t half)
{
    // t = x * a + y * b

    const uint16x8_t ta = vmulq_u16(x, a);
    const uint16x8_t tb = vmulq_u16(y, b);

    return qvdiv_255_u16(vaddq_u16(ta, tb), half);
}

static inline uint16x8_t qvsource_over_u16(uint16x8_t src16, uint16x8_t dst16, uint16x8_t half, uint16x8_t full)
{
    const uint16x4_t alpha16_high = vdup_lane_u16(vget_high_u16(src16), 3);
    const uint16x4_t alpha16_low = vdup_lane_u16(vget_low_u16(src16), 3);

    const uint16x8_t alpha16 = vsubq_u16(full, vcombine_u16(alpha16_low, alpha16_high));

    return vaddq_u16(src16, qvbyte_mul_u16(dst16, alpha16, half));
}

extern "C" void
pixman_composite_over_8888_0565_asm_neon (int32_t   w,
                                          int32_t   h,
                                          uint16_t *dst,
                                          int32_t   dst_stride,
                                          uint32_t *src,
                                          int32_t   src_stride);

extern "C" void
pixman_composite_over_8888_8888_asm_neon (int32_t   w,
                                          int32_t   h,
                                          uint32_t *dst,
                                          int32_t   dst_stride,
                                          uint32_t *src,
                                          int32_t   src_stride);

extern "C" void
pixman_composite_src_0565_8888_asm_neon (int32_t   w,
                                         int32_t   h,
                                         uint32_t *dst,
                                         int32_t   dst_stride,
                                         uint16_t *src,
                                         int32_t   src_stride);

extern "C" void
pixman_composite_over_n_8_0565_asm_neon (int32_t    w,
                                         int32_t    h,
                                         uint16_t  *dst,
                                         int32_t    dst_stride,
                                         uint32_t   src,
                                         int32_t    unused,
                                         uint8_t   *mask,
                                         int32_t    mask_stride);

extern "C" void
pixman_composite_scanline_over_asm_neon (int32_t         w,
                                         const uint32_t *dst,
                                         const uint32_t *src);

extern "C" void
pixman_composite_src_0565_0565_asm_neon (int32_t   w,
                                         int32_t   h,
                                         uint16_t *dst,
                                         int32_t   dst_stride,
                                         uint16_t *src,
                                         int32_t   src_stride);

// qblendfunctions.cpp
void qt_blend_argb32_on_rgb16_const_alpha(uchar *destPixels, int dbpl,
                                          const uchar *srcPixels, int sbpl,
                                          int w, int h,
                                          int const_alpha);

void qt_blend_rgb16_on_argb32_neon(uchar *destPixels, int dbpl,
                                   const uchar *srcPixels, int sbpl,
                                   int w, int h,
                                   int const_alpha)
{
    dbpl /= 4;
    sbpl /= 2;

    quint32 *dst = (quint32 *) destPixels;
    quint16 *src = (quint16 *) srcPixels;

    if (const_alpha != 256) {
        quint8 a = (255 * const_alpha) >> 8;
        quint8 ia = 255 - a;

        while (h--) {
            for (int x=0; x<w; ++x)
                dst[x] = INTERPOLATE_PIXEL_255(qt_colorConvert(src[x], dst[x]), a, dst[x], ia);
            dst += dbpl;
            src += sbpl;
        }
        return;
    }

    pixman_composite_src_0565_8888_asm_neon(w, h, dst, dbpl, src, sbpl);
}

// qblendfunctions.cpp
void qt_blend_rgb16_on_rgb16(uchar *dst, int dbpl,
                             const uchar *src, int sbpl,
                             int w, int h,
                             int const_alpha);

template <int N>
static inline void scanLineBlit16(quint16 *dst, quint16 *src, int dstride)
{
    if (N >= 2) {
        ((quint32 *)dst)[0] = ((quint32 *)src)[0];
        __builtin_prefetch(dst + dstride, 1, 0);
    }
    for (int i = 1; i < N/2; ++i)
        ((quint32 *)dst)[i] = ((quint32 *)src)[i];
    if (N & 1)
        dst[N-1] = src[N-1];
}

template <int Width>
static inline void blockBlit16(quint16 *dst, quint16 *src, int dstride, int sstride, int h)
{
    union {
        quintptr address;
        quint16 *pointer;
    } u;

    u.pointer = dst;

    if (u.address & 2) {
        while (h--) {
            // align dst
            dst[0] = src[0];
            if (Width > 1)
                scanLineBlit16<Width-1>(dst + 1, src + 1, dstride);
            dst += dstride;
            src += sstride;
        }
    } else {
        while (h--) {
            scanLineBlit16<Width>(dst, src, dstride);

            dst += dstride;
            src += sstride;
        }
    }
}

void qt_blend_rgb16_on_rgb16_neon(uchar *destPixels, int dbpl,
                                  const uchar *srcPixels, int sbpl,
                                  int w, int h,
                                  int const_alpha)
{
    // testing show that the default memcpy is faster for widths 150 and up
    if (const_alpha != 256 || w >= 150) {
        qt_blend_rgb16_on_rgb16(destPixels, dbpl, srcPixels, sbpl, w, h, const_alpha);
        return;
    }

    int dstride = dbpl / 2;
    int sstride = sbpl / 2;

    quint16 *dst = (quint16 *) destPixels;
    quint16 *src = (quint16 *) srcPixels;

    switch (w) {
#define BLOCKBLIT(n) case n: blockBlit16<n>(dst, src, dstride, sstride, h); return;
    BLOCKBLIT(1);
    BLOCKBLIT(2);
    BLOCKBLIT(3);
    BLOCKBLIT(4);
    BLOCKBLIT(5);
    BLOCKBLIT(6);
    BLOCKBLIT(7);
    BLOCKBLIT(8);
    BLOCKBLIT(9);
    BLOCKBLIT(10);
    BLOCKBLIT(11);
    BLOCKBLIT(12);
    BLOCKBLIT(13);
    BLOCKBLIT(14);
    BLOCKBLIT(15);
#undef BLOCKBLIT
    default:
        break;
    }

    pixman_composite_src_0565_0565_asm_neon (w, h, dst, dstride, src, sstride);
}

extern "C" void blend_8_pixels_argb32_on_rgb16_neon(quint16 *dst, const quint32 *src, int const_alpha);

void qt_blend_argb32_on_rgb16_neon(uchar *destPixels, int dbpl,
                                   const uchar *srcPixels, int sbpl,
                                   int w, int h,
                                   int const_alpha)
{
    quint16 *dst = (quint16 *) destPixels;
    quint32 *src = (quint32 *) srcPixels;

    if (const_alpha != 256) {
        for (int y=0; y<h; ++y) {
            int i = 0;
            for (; i < w-7; i += 8)
                blend_8_pixels_argb32_on_rgb16_neon(&dst[i], &src[i], const_alpha);

            if (i < w) {
                int tail = w - i;

                quint16 dstBuffer[8];
                quint32 srcBuffer[8];

                for (int j = 0; j < tail; ++j) {
                    dstBuffer[j] = dst[i + j];
                    srcBuffer[j] = src[i + j];
                }

                blend_8_pixels_argb32_on_rgb16_neon(dstBuffer, srcBuffer, const_alpha);

                for (int j = 0; j < tail; ++j)
                    dst[i + j] = dstBuffer[j];
            }

            dst = (quint16 *)(((uchar *) dst) + dbpl);
            src = (quint32 *)(((uchar *) src) + sbpl);
        }
        return;
    }

    pixman_composite_over_8888_0565_asm_neon(w, h, dst, dbpl / 2, src, sbpl / 4);
}

void qt_blend_argb32_on_argb32_scanline_neon(uint *dest, const uint *src, int length, uint const_alpha)
{
    if (const_alpha == 255) {
        pixman_composite_scanline_over_asm_neon(length, dest, src);
    } else {
        qt_blend_argb32_on_argb32_neon((uchar *)dest, 4 * length, (uchar *)src, 4 * length, length, 1, (const_alpha * 256) / 255);
    }
}

void qt_blend_argb32_on_argb32_neon(uchar *destPixels, int dbpl,
                                    const uchar *srcPixels, int sbpl,
                                    int w, int h,
                                    int const_alpha)
{
    const uint *src = (const uint *) srcPixels;
    uint *dst = (uint *) destPixels;
    uint16x8_t half = vdupq_n_u16(0x80);
    uint16x8_t full = vdupq_n_u16(0xff);
    if (const_alpha == 256) {
        pixman_composite_over_8888_8888_asm_neon(w, h, (uint32_t *)destPixels, dbpl / 4, (uint32_t *)srcPixels, sbpl / 4);
    } else if (const_alpha != 0) {
        const_alpha = (const_alpha * 255) >> 8;
        uint16x8_t const_alpha16 = vdupq_n_u16(const_alpha);
        for (int y = 0; y < h; ++y) {
            int x = 0;
            for (; x < w-3; x += 4) {
                if (src[x] | src[x+1] | src[x+2] | src[x+3]) {
                    uint32x4_t src32 = vld1q_u32((uint32_t *)&src[x]);
                    uint32x4_t dst32 = vld1q_u32((uint32_t *)&dst[x]);

                    const uint8x16_t src8 = vreinterpretq_u8_u32(src32);
                    const uint8x16_t dst8 = vreinterpretq_u8_u32(dst32);

                    const uint8x8_t src8_low = vget_low_u8(src8);
                    const uint8x8_t dst8_low = vget_low_u8(dst8);

                    const uint8x8_t src8_high = vget_high_u8(src8);
                    const uint8x8_t dst8_high = vget_high_u8(dst8);

                    const uint16x8_t src16_low = vmovl_u8(src8_low);
                    const uint16x8_t dst16_low = vmovl_u8(dst8_low);

                    const uint16x8_t src16_high = vmovl_u8(src8_high);
                    const uint16x8_t dst16_high = vmovl_u8(dst8_high);

                    const uint16x8_t srcalpha16_low = qvbyte_mul_u16(src16_low, const_alpha16, half);
                    const uint16x8_t srcalpha16_high = qvbyte_mul_u16(src16_high, const_alpha16, half);

                    const uint16x8_t result16_low = qvsource_over_u16(srcalpha16_low, dst16_low, half, full);
                    const uint16x8_t result16_high = qvsource_over_u16(srcalpha16_high, dst16_high, half, full);

                    const uint32x2_t result32_low = vreinterpret_u32_u8(vmovn_u16(result16_low));
                    const uint32x2_t result32_high = vreinterpret_u32_u8(vmovn_u16(result16_high));

                    vst1q_u32((uint32_t *)&dst[x], vcombine_u32(result32_low, result32_high));
                }
            }
            for (; x<w; ++x) {
                uint s = src[x];
                if (s != 0) {
                    s = BYTE_MUL(s, const_alpha);
                    dst[x] = s + BYTE_MUL(dst[x], qAlpha(~s));
                }
            }
            dst = (quint32 *)(((uchar *) dst) + dbpl);
            src = (const quint32 *)(((const uchar *) src) + sbpl);
        }
    }
}

// qblendfunctions.cpp
void qt_blend_rgb32_on_rgb32(uchar *destPixels, int dbpl,
                             const uchar *srcPixels, int sbpl,
                             int w, int h,
                             int const_alpha);

void qt_blend_rgb32_on_rgb32_neon(uchar *destPixels, int dbpl,
                                  const uchar *srcPixels, int sbpl,
                                  int w, int h,
                                  int const_alpha)
{
    if (const_alpha != 256) {
        if (const_alpha != 0) {
            const uint *src = (const uint *) srcPixels;
            uint *dst = (uint *) destPixels;
            uint16x8_t half = vdupq_n_u16(0x80);
            const_alpha = (const_alpha * 255) >> 8;
            int one_minus_const_alpha = 255 - const_alpha;
            uint16x8_t const_alpha16 = vdupq_n_u16(const_alpha);
            uint16x8_t one_minus_const_alpha16 = vdupq_n_u16(255 - const_alpha);
            for (int y = 0; y < h; ++y) {
                int x = 0;
                for (; x < w-3; x += 4) {
                    uint32x4_t src32 = vld1q_u32((uint32_t *)&src[x]);
                    uint32x4_t dst32 = vld1q_u32((uint32_t *)&dst[x]);

                    const uint8x16_t src8 = vreinterpretq_u8_u32(src32);
                    const uint8x16_t dst8 = vreinterpretq_u8_u32(dst32);

                    const uint8x8_t src8_low = vget_low_u8(src8);
                    const uint8x8_t dst8_low = vget_low_u8(dst8);

                    const uint8x8_t src8_high = vget_high_u8(src8);
                    const uint8x8_t dst8_high = vget_high_u8(dst8);

                    const uint16x8_t src16_low = vmovl_u8(src8_low);
                    const uint16x8_t dst16_low = vmovl_u8(dst8_low);

                    const uint16x8_t src16_high = vmovl_u8(src8_high);
                    const uint16x8_t dst16_high = vmovl_u8(dst8_high);

                    const uint16x8_t result16_low = qvinterpolate_pixel_255(src16_low, const_alpha16, dst16_low, one_minus_const_alpha16, half);
                    const uint16x8_t result16_high = qvinterpolate_pixel_255(src16_high, const_alpha16, dst16_high, one_minus_const_alpha16, half);

                    const uint32x2_t result32_low = vreinterpret_u32_u8(vmovn_u16(result16_low));
                    const uint32x2_t result32_high = vreinterpret_u32_u8(vmovn_u16(result16_high));

                    vst1q_u32((uint32_t *)&dst[x], vcombine_u32(result32_low, result32_high));
                }
                for (; x<w; ++x) {
                    uint s = src[x];
                    s = BYTE_MUL(s, const_alpha);
                    dst[x] = INTERPOLATE_PIXEL_255(src[x], const_alpha, dst[x], one_minus_const_alpha);
                }
                dst = (quint32 *)(((uchar *) dst) + dbpl);
                src = (const quint32 *)(((const uchar *) src) + sbpl);
            }
        }
    } else {
        qt_blend_rgb32_on_rgb32(destPixels, dbpl, srcPixels, sbpl, w, h, const_alpha);
    }
}

void qt_alphamapblit_quint16_neon(QRasterBuffer *rasterBuffer,
                                  int x, int y, quint32 color,
                                  const uchar *bitmap,
                                  int mapWidth, int mapHeight, int mapStride,
                                  const QClipData *)
{
    quint16 *dest = reinterpret_cast<quint16*>(rasterBuffer->scanLine(y)) + x;
    const int destStride = rasterBuffer->bytesPerLine() / sizeof(quint16);

    uchar *mask = const_cast<uchar *>(bitmap);

    pixman_composite_over_n_8_0565_asm_neon(mapWidth, mapHeight, dest, destStride, color, 0, mask, mapStride);
}

extern "C" void blend_8_pixels_rgb16_on_rgb16_neon(quint16 *dst, const quint16 *src, int const_alpha);

template <typename SRC, typename BlendFunc>
struct Blend_on_RGB16_SourceAndConstAlpha_Neon {
    Blend_on_RGB16_SourceAndConstAlpha_Neon(BlendFunc blender, int const_alpha)
        : m_index(0)
        , m_blender(blender)
        , m_const_alpha(const_alpha)
    {
    }

    inline void write(quint16 *dst, quint32 src)
    {
        srcBuffer[m_index++] = src;

        if (m_index == 8) {
            m_blender(dst - 7, srcBuffer, m_const_alpha);
            m_index = 0;
        }
    }

    inline void flush(quint16 *dst)
    {
        if (m_index > 0) {
            quint16 dstBuffer[8];
            for (int i = 0; i < m_index; ++i)
                dstBuffer[i] = dst[i - m_index];

            m_blender(dstBuffer, srcBuffer, m_const_alpha);

            for (int i = 0; i < m_index; ++i)
                dst[i - m_index] = dstBuffer[i];

            m_index = 0;
        }
    }

    SRC srcBuffer[8];

    int m_index;
    BlendFunc m_blender;
    int m_const_alpha;
};

template <typename SRC, typename BlendFunc>
Blend_on_RGB16_SourceAndConstAlpha_Neon<SRC, BlendFunc>
Blend_on_RGB16_SourceAndConstAlpha_Neon_create(BlendFunc blender, int const_alpha)
{
    return Blend_on_RGB16_SourceAndConstAlpha_Neon<SRC, BlendFunc>(blender, const_alpha);
}

void qt_scale_image_argb32_on_rgb16_neon(uchar *destPixels, int dbpl,
                                         const uchar *srcPixels, int sbpl,
                                         const QRectF &targetRect,
                                         const QRectF &sourceRect,
                                         const QRect &clip,
                                         int const_alpha)
{
    if (const_alpha == 0)
        return;

    qt_scale_image_16bit<quint32>(destPixels, dbpl, srcPixels, sbpl, targetRect, sourceRect, clip,
        Blend_on_RGB16_SourceAndConstAlpha_Neon_create<quint32>(blend_8_pixels_argb32_on_rgb16_neon, const_alpha));
}

void qt_scale_image_rgb16_on_rgb16(uchar *destPixels, int dbpl,
                                   const uchar *srcPixels, int sbpl,
                                   const QRectF &targetRect,
                                   const QRectF &sourceRect,
                                   const QRect &clip,
                                   int const_alpha);

void qt_scale_image_rgb16_on_rgb16_neon(uchar *destPixels, int dbpl,
                                        const uchar *srcPixels, int sbpl,
                                        const QRectF &targetRect,
                                        const QRectF &sourceRect,
                                        const QRect &clip,
                                        int const_alpha)
{
    if (const_alpha == 0)
        return;

    if (const_alpha == 256) {
        qt_scale_image_rgb16_on_rgb16(destPixels, dbpl, srcPixels, sbpl, targetRect, sourceRect, clip, const_alpha);
        return;
    }

    qt_scale_image_16bit<quint16>(destPixels, dbpl, srcPixels, sbpl, targetRect, sourceRect, clip,
        Blend_on_RGB16_SourceAndConstAlpha_Neon_create<quint16>(blend_8_pixels_rgb16_on_rgb16_neon, const_alpha));
}

extern void qt_transform_image_rgb16_on_rgb16(uchar *destPixels, int dbpl,
                                              const uchar *srcPixels, int sbpl,
                                              const QRectF &targetRect,
                                              const QRectF &sourceRect,
                                              const QRect &clip,
                                              const QTransform &targetRectTransform,
                                              int const_alpha);

void qt_transform_image_rgb16_on_rgb16_neon(uchar *destPixels, int dbpl,
                                            const uchar *srcPixels, int sbpl,
                                            const QRectF &targetRect,
                                            const QRectF &sourceRect,
                                            const QRect &clip,
                                            const QTransform &targetRectTransform,
                                            int const_alpha)
{
    if (const_alpha == 0)
        return;

    if (const_alpha == 256) {
        qt_transform_image_rgb16_on_rgb16(destPixels, dbpl, srcPixels, sbpl, targetRect, sourceRect, clip, targetRectTransform, const_alpha);
        return;
    }

    qt_transform_image(reinterpret_cast<quint16 *>(destPixels), dbpl,
                       reinterpret_cast<const quint16 *>(srcPixels), sbpl, targetRect, sourceRect, clip, targetRectTransform,
        Blend_on_RGB16_SourceAndConstAlpha_Neon_create<quint16>(blend_8_pixels_rgb16_on_rgb16_neon, const_alpha));
}

void qt_transform_image_argb32_on_rgb16_neon(uchar *destPixels, int dbpl,
                                             const uchar *srcPixels, int sbpl,
                                             const QRectF &targetRect,
                                             const QRectF &sourceRect,
                                             const QRect &clip,
                                             const QTransform &targetRectTransform,
                                             int const_alpha)
{
    if (const_alpha == 0)
        return;

    qt_transform_image(reinterpret_cast<quint16 *>(destPixels), dbpl,
                       reinterpret_cast<const quint32 *>(srcPixels), sbpl, targetRect, sourceRect, clip, targetRectTransform,
        Blend_on_RGB16_SourceAndConstAlpha_Neon_create<quint32>(blend_8_pixels_argb32_on_rgb16_neon, const_alpha));
}

static inline void convert_8_pixels_rgb16_to_argb32(quint32 *dst, const quint16 *src)
{
    asm volatile (
        "vld1.16     { d0, d1 }, [%[SRC]]\n\t"

        /* convert 8 r5g6b5 pixel data from {d0, d1} to planar 8-bit format
           and put data into d4 - red, d3 - green, d2 - blue */
        "vshrn.u16   d4,  q0,  #8\n\t"
        "vshrn.u16   d3,  q0,  #3\n\t"
        "vsli.u16    q0,  q0,  #5\n\t"
        "vsri.u8     d4,  d4,  #5\n\t"
        "vsri.u8     d3,  d3,  #6\n\t"
        "vshrn.u16   d2,  q0,  #2\n\t"

        /* fill d5 - alpha with 0xff */
        "mov         r2, #255\n\t"
        "vdup.8      d5, r2\n\t"

        "vst4.8      { d2, d3, d4, d5 }, [%[DST]]"
        : : [DST]"r" (dst), [SRC]"r" (src)
        : "memory", "r2", "d0", "d1", "d2", "d3", "d4", "d5"
    );
}

uint * QT_FASTCALL qt_destFetchRGB16_neon(uint *buffer, QRasterBuffer *rasterBuffer, int x, int y, int length)
{
    const ushort *data = (const ushort *)rasterBuffer->scanLine(y) + x;

    int i = 0;
    for (; i < length - 7; i += 8)
        convert_8_pixels_rgb16_to_argb32(&buffer[i], &data[i]);

    if (i < length) {
        quint16 srcBuffer[8];
        quint32 dstBuffer[8];

        int tail = length - i;
        for (int j = 0; j < tail; ++j)
            srcBuffer[j] = data[i + j];

        convert_8_pixels_rgb16_to_argb32(dstBuffer, srcBuffer);

        for (int j = 0; j < tail; ++j)
            buffer[i + j] = dstBuffer[j];
    }

    return buffer;
}

static inline void convert_8_pixels_argb32_to_rgb16(quint16 *dst, const quint32 *src)
{
    asm volatile (
        "vld4.8      { d0, d1, d2, d3 }, [%[SRC]]\n\t"

        /* convert to r5g6b5 and store it into {d28, d29} */
        "vshll.u8    q14, d2, #8\n\t"
        "vshll.u8    q8,  d1, #8\n\t"
        "vshll.u8    q9,  d0, #8\n\t"
        "vsri.u16    q14, q8, #5\n\t"
        "vsri.u16    q14, q9, #11\n\t"

        "vst1.16     { d28, d29 }, [%[DST]]"
        : : [DST]"r" (dst), [SRC]"r" (src)
        : "memory", "d0", "d1", "d2", "d3", "d16", "d17", "d18", "d19", "d28", "d29"
    );
}

void QT_FASTCALL qt_destStoreRGB16_neon(QRasterBuffer *rasterBuffer, int x, int y, const uint *buffer, int length)
{
    quint16 *data = (quint16*)rasterBuffer->scanLine(y) + x;

    int i = 0;
    for (; i < length - 7; i += 8)
        convert_8_pixels_argb32_to_rgb16(&data[i], &buffer[i]);

    if (i < length) {
        quint32 srcBuffer[8];
        quint16 dstBuffer[8];

        int tail = length - i;
        for (int j = 0; j < tail; ++j)
            srcBuffer[j] = buffer[i + j];

        convert_8_pixels_argb32_to_rgb16(dstBuffer, srcBuffer);

        for (int j = 0; j < tail; ++j)
            data[i + j] = dstBuffer[j];
    }
}

void QT_FASTCALL comp_func_solid_SourceOver_neon(uint *destPixels, int length, uint color, uint const_alpha)
{
    if ((const_alpha & qAlpha(color)) == 255) {
        QT_MEMFILL_UINT(destPixels, length, color);
    } else {
        if (const_alpha != 255)
            color = BYTE_MUL(color, const_alpha);

        const quint32 minusAlphaOfColor = qAlpha(~color);
        int x = 0;

        uint32_t *dst = (uint32_t *) destPixels;
        const uint32x4_t colorVector = vdupq_n_u32(color);
        uint16x8_t half = vdupq_n_u16(0x80);
        const uint16x8_t minusAlphaOfColorVector = vdupq_n_u16(minusAlphaOfColor);

        for (; x < length-3; x += 4) {
            uint32x4_t dstVector = vld1q_u32(&dst[x]);

            const uint8x16_t dst8 = vreinterpretq_u8_u32(dstVector);

            const uint8x8_t dst8_low = vget_low_u8(dst8);
            const uint8x8_t dst8_high = vget_high_u8(dst8);

            const uint16x8_t dst16_low = vmovl_u8(dst8_low);
            const uint16x8_t dst16_high = vmovl_u8(dst8_high);

            const uint16x8_t result16_low = qvbyte_mul_u16(dst16_low, minusAlphaOfColorVector, half);
            const uint16x8_t result16_high = qvbyte_mul_u16(dst16_high, minusAlphaOfColorVector, half);

            const uint32x2_t result32_low = vreinterpret_u32_u8(vmovn_u16(result16_low));
            const uint32x2_t result32_high = vreinterpret_u32_u8(vmovn_u16(result16_high));

            uint32x4_t blendedPixels = vcombine_u32(result32_low, result32_high);
            uint32x4_t colorPlusBlendedPixels = vaddq_u32(colorVector, blendedPixels);
            vst1q_u32(&dst[x], colorPlusBlendedPixels);
        }

        for (;x < length; ++x)
            destPixels[x] = color + BYTE_MUL(destPixels[x], minusAlphaOfColor);
    }
}

void QT_FASTCALL comp_func_Plus_neon(uint *dst, const uint *src, int length, uint const_alpha)
{
    if (const_alpha == 255) {
        uint *const end = dst + length;
        uint *const neonEnd = end - 3;

        while (dst < neonEnd) {
            asm volatile (
                "vld2.8     { d0, d1 }, [%[SRC]] !\n\t"
                "vld2.8     { d2, d3 }, [%[DST]]\n\t"
                "vqadd.u8 q0, q0, q1\n\t"
                "vst2.8     { d0, d1 }, [%[DST]] !\n\t"
                : [DST]"+r" (dst), [SRC]"+r" (src)
                :
                : "memory", "d0", "d1", "d2", "d3", "q0", "q1"
            );
        }

        while (dst != end) {
            *dst = comp_func_Plus_one_pixel(*dst, *src);
            ++dst;
            ++src;
        }
    } else {
        int x = 0;
        const int one_minus_const_alpha = 255 - const_alpha;
        const uint16x8_t constAlphaVector = vdupq_n_u16(const_alpha);
        const uint16x8_t oneMinusconstAlphaVector = vdupq_n_u16(one_minus_const_alpha);

        const uint16x8_t half = vdupq_n_u16(0x80);
        for (; x < length - 3; x += 4) {
            const uint32x4_t src32 = vld1q_u32((uint32_t *)&src[x]);
            const uint8x16_t src8 = vreinterpretq_u8_u32(src32);
            uint8x16_t dst8 = vld1q_u8((uint8_t *)&dst[x]);
            uint8x16_t result = vqaddq_u8(dst8, src8);

            uint16x8_t result_low = vmovl_u8(vget_low_u8(result));
            uint16x8_t result_high = vmovl_u8(vget_high_u8(result));

            uint16x8_t dst_low = vmovl_u8(vget_low_u8(dst8));
            uint16x8_t dst_high = vmovl_u8(vget_high_u8(dst8));

            result_low = qvinterpolate_pixel_255(result_low, constAlphaVector, dst_low, oneMinusconstAlphaVector, half);
            result_high = qvinterpolate_pixel_255(result_high, constAlphaVector, dst_high, oneMinusconstAlphaVector, half);

            const uint32x2_t result32_low = vreinterpret_u32_u8(vmovn_u16(result_low));
            const uint32x2_t result32_high = vreinterpret_u32_u8(vmovn_u16(result_high));
            vst1q_u32((uint32_t *)&dst[x], vcombine_u32(result32_low, result32_high));
        }

        for (; x < length; ++x)
            dst[x] = comp_func_Plus_one_pixel_const_alpha(dst[x], src[x], const_alpha, one_minus_const_alpha);
    }
}

static const int tileSize = 32;

extern "C" void qt_rotate90_16_neon(quint16 *dst, const quint16 *src, int sstride, int dstride, int count);

void qt_memrotate90_16_neon(const uchar *srcPixels, int w, int h, int sstride, uchar *destPixels, int dstride)
{
    const ushort *src = (const ushort *)srcPixels;
    ushort *dest = (ushort *)destPixels;

    sstride /= sizeof(ushort);
    dstride /= sizeof(ushort);

    const int pack = sizeof(quint32) / sizeof(ushort);
    const int unaligned =
        qMin(uint((quintptr(dest) & (sizeof(quint32)-1)) / sizeof(ushort)), uint(h));
    const int restX = w % tileSize;
    const int restY = (h - unaligned) % tileSize;
    const int unoptimizedY = restY % pack;
    const int numTilesX = w / tileSize + (restX > 0);
    const int numTilesY = (h - unaligned) / tileSize + (restY >= pack);

    for (int tx = 0; tx < numTilesX; ++tx) {
        const int startx = w - tx * tileSize - 1;
        const int stopx = qMax(startx - tileSize, 0);

        if (unaligned) {
            for (int x = startx; x >= stopx; --x) {
                ushort *d = dest + (w - x - 1) * dstride;
                for (int y = 0; y < unaligned; ++y) {
                    *d++ = src[y * sstride + x];
                }
            }
        }

        for (int ty = 0; ty < numTilesY; ++ty) {
            const int starty = ty * tileSize + unaligned;
            const int stopy = qMin(starty + tileSize, h - unoptimizedY);

            int x = startx;
            // qt_rotate90_16_neon writes to eight rows, four pixels at a time
            for (; x >= stopx + 7; x -= 8) {
                ushort *d = dest + (w - x - 1) * dstride + starty;
                const ushort *s = &src[starty * sstride + x - 7];
                qt_rotate90_16_neon(d, s, sstride * 2, dstride * 2, stopy - starty);
            }

            for (; x >= stopx; --x) {
                quint32 *d = reinterpret_cast<quint32*>(dest + (w - x - 1) * dstride + starty);
                for (int y = starty; y < stopy; y += pack) {
                    quint32 c = src[y * sstride + x];
                    for (int i = 1; i < pack; ++i) {
                        const int shift = (sizeof(int) * 8 / pack * i);
                        const ushort color = src[(y + i) * sstride + x];
                        c |= color << shift;
                    }
                    *d++ = c;
                }
            }
        }

        if (unoptimizedY) {
            const int starty = h - unoptimizedY;
            for (int x = startx; x >= stopx; --x) {
                ushort *d = dest + (w - x - 1) * dstride + starty;
                for (int y = starty; y < h; ++y) {
                    *d++ = src[y * sstride + x];
                }
            }
        }
    }
}

extern "C" void qt_rotate270_16_neon(quint16 *dst, const quint16 *src, int sstride, int dstride, int count);

void qt_memrotate270_16_neon(const uchar *srcPixels, int w, int h,
                             int sstride,
                             uchar *destPixels, int dstride)
{
    const ushort *src = (const ushort *)srcPixels;
    ushort *dest = (ushort *)destPixels;

    sstride /= sizeof(ushort);
    dstride /= sizeof(ushort);

    const int pack = sizeof(quint32) / sizeof(ushort);
    const int unaligned =
        qMin(uint((long(dest) & (sizeof(quint32)-1)) / sizeof(ushort)), uint(h));
    const int restX = w % tileSize;
    const int restY = (h - unaligned) % tileSize;
    const int unoptimizedY = restY % pack;
    const int numTilesX = w / tileSize + (restX > 0);
    const int numTilesY = (h - unaligned) / tileSize + (restY >= pack);

    for (int tx = 0; tx < numTilesX; ++tx) {
        const int startx = tx * tileSize;
        const int stopx = qMin(startx + tileSize, w);

        if (unaligned) {
            for (int x = startx; x < stopx; ++x) {
                ushort *d = dest + x * dstride;
                for (int y = h - 1; y >= h - unaligned; --y) {
                    *d++ = src[y * sstride + x];
                }
            }
        }

        for (int ty = 0; ty < numTilesY; ++ty) {
            const int starty = h - 1 - unaligned - ty * tileSize;
            const int stopy = qMax(starty - tileSize, unoptimizedY);

            int x = startx;
            // qt_rotate90_16_neon writes to eight rows, four pixels at a time
            for (; x < stopx - 7; x += 8) {
                ushort *d = dest + x * dstride + h - 1 - starty;
                const ushort *s = &src[starty * sstride + x];
                qt_rotate90_16_neon(d + 7 * dstride, s, -sstride * 2, -dstride * 2, starty - stopy);
            }

            for (; x < stopx; ++x) {
                quint32 *d = reinterpret_cast<quint32*>(dest + x * dstride
                                                        + h - 1 - starty);
                for (int y = starty; y > stopy; y -= pack) {
                    quint32 c = src[y * sstride + x];
                    for (int i = 1; i < pack; ++i) {
                        const int shift = (sizeof(int) * 8 / pack * i);
                        const ushort color = src[(y - i) * sstride + x];
                        c |= color << shift;
                    }
                    *d++ = c;
                }
            }
        }
        if (unoptimizedY) {
            const int starty = unoptimizedY - 1;
            for (int x = startx; x < stopx; ++x) {
                ushort *d = dest + x * dstride + h - 1 - starty;
                for (int y = starty; y >= 0; --y) {
                    *d++ = src[y * sstride + x];
                }
            }
        }
    }
}

class QSimdNeon
{
public:
    typedef int32x4_t Int32x4;
    typedef float32x4_t Float32x4;

    union Vect_buffer_i { Int32x4 v; int i[4]; };
    union Vect_buffer_f { Float32x4 v; float f[4]; };

    static inline Float32x4 v_dup(float x) { return vdupq_n_f32(x); }
    static inline Int32x4 v_dup(int x) { return vdupq_n_s32(x); }
    static inline Int32x4 v_dup(uint x) { return vdupq_n_s32(x); }

    static inline Float32x4 v_add(Float32x4 a, Float32x4 b) { return vaddq_f32(a, b); }
    static inline Int32x4 v_add(Int32x4 a, Int32x4 b) { return vaddq_s32(a, b); }

    static inline Float32x4 v_max(Float32x4 a, Float32x4 b) { return vmaxq_f32(a, b); }
    static inline Float32x4 v_min(Float32x4 a, Float32x4 b) { return vminq_f32(a, b); }
    static inline Int32x4 v_min_16(Int32x4 a, Int32x4 b) { return vminq_s32(a, b); }

    static inline Int32x4 v_and(Int32x4 a, Int32x4 b) { return vandq_s32(a, b); }

    static inline Float32x4 v_sub(Float32x4 a, Float32x4 b) { return vsubq_f32(a, b); }
    static inline Int32x4 v_sub(Int32x4 a, Int32x4 b) { return vsubq_s32(a, b); }

    static inline Float32x4 v_mul(Float32x4 a, Float32x4 b) { return vmulq_f32(a, b); }

    static inline Float32x4 v_sqrt(Float32x4 x) { Float32x4 y = vrsqrteq_f32(x); y = vmulq_f32(y, vrsqrtsq_f32(x, vmulq_f32(y, y))); return vmulq_f32(x, y); }

    static inline Int32x4 v_toInt(Float32x4 x) { return vcvtq_s32_f32(x); }

    static inline Int32x4 v_greaterOrEqual(Float32x4 a, Float32x4 b) { return vreinterpretq_s32_u32(vcgeq_f32(a, b)); }
};

const uint * QT_FASTCALL qt_fetch_radial_gradient_neon(uint *buffer, const Operator *op, const QSpanData *data,
                                                       int y, int x, int length)
{
    return qt_fetch_radial_gradient_template<QRadialFetchSimd<QSimdNeon> >(buffer, op, data, y, x, length);
}

QT_END_NAMESPACE

#endif // QT_HAVE_NEON