1/****************************************************************************
2**
3** Copyright (C) 2016 The Qt Company Ltd.
4** Contact: https://www.qt.io/licensing/
5**
6** This file is part of the QtGui module of the Qt Toolkit.
7**
8** $QT_BEGIN_LICENSE:LGPL$
9** Commercial License Usage
10** Licensees holding valid commercial Qt licenses may use this file in
11** accordance with the commercial license agreement provided with the
12** Software or, alternatively, in accordance with the terms contained in
13** a written agreement between you and The Qt Company. For licensing terms
14** and conditions see https://www.qt.io/terms-conditions. For further
15** information use the contact form at https://www.qt.io/contact-us.
16**
17** GNU Lesser General Public License Usage
18** Alternatively, this file may be used under the terms of the GNU Lesser
19** General Public License version 3 as published by the Free Software
20** Foundation and appearing in the file LICENSE.LGPL3 included in the
21** packaging of this file. Please review the following information to
22** ensure the GNU Lesser General Public License version 3 requirements
23** will be met: https://www.gnu.org/licenses/lgpl-3.0.html.
24**
25** GNU General Public License Usage
26** Alternatively, this file may be used under the terms of the GNU
27** General Public License version 2.0 or (at your option) the GNU General
28** Public license version 3 or any later version approved by the KDE Free
29** Qt Foundation. The licenses are as published by the Free Software
30** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3
31** included in the packaging of this file. Please review the following
32** information to ensure the GNU General Public License requirements will
33** be met: https://www.gnu.org/licenses/gpl-2.0.html and
34** https://www.gnu.org/licenses/gpl-3.0.html.
35**
36** $QT_END_LICENSE$
37**
38****************************************************************************/
39
40#ifndef QDRAWHELPER_P_H
41#define QDRAWHELPER_P_H
42
43//
44// W A R N I N G
45// -------------
46//
47// This file is not part of the Qt API. It exists purely as an
48// implementation detail. This header file may change from version to
49// version without notice, or even be removed.
50//
51// We mean it.
52//
53
54#include <QtGui/private/qtguiglobal_p.h>
55#include "QtCore/qmath.h"
56#include "QtGui/qcolor.h"
57#include "QtGui/qpainter.h"
58#include "QtGui/qimage.h"
59#include "QtGui/qrgba64.h"
60#ifndef QT_FT_BEGIN_HEADER
61#define QT_FT_BEGIN_HEADER
62#define QT_FT_END_HEADER
63#endif
64#include "private/qpixellayout_p.h"
65#include "private/qrasterdefs_p.h"
66#include <private/qsimd_p.h>
67
68#include <QtCore/qsharedpointer.h>
69
70QT_BEGIN_NAMESPACE
71
72#if defined(Q_CC_GNU)
73# define Q_DECL_RESTRICT __restrict__
74# if defined(Q_PROCESSOR_X86_32) && defined(Q_CC_GNU) && !defined(Q_CC_CLANG) && !defined(Q_CC_INTEL)
75# define Q_DECL_VECTORCALL __attribute__((sseregparm,regparm(3)))
76# else
77# define Q_DECL_VECTORCALL
78# endif
79#elif defined(Q_CC_MSVC)
80# define Q_DECL_RESTRICT __restrict
81# define Q_DECL_VECTORCALL __vectorcall
82#else
83# define Q_DECL_RESTRICT
84# define Q_DECL_VECTORCALL
85#endif
86
87static const uint AMASK = 0xff000000;
88static const uint RMASK = 0x00ff0000;
89static const uint GMASK = 0x0000ff00;
90static const uint BMASK = 0x000000ff;
91
92/*******************************************************************************
93 * QSpan
94 *
95 * duplicate definition of FT_Span
96 */
97typedef QT_FT_Span QSpan;
98
99struct QSolidData;
100struct QTextureData;
101struct QGradientData;
102struct QLinearGradientData;
103struct QRadialGradientData;
104struct QConicalGradientData;
105struct QSpanData;
106class QGradient;
107class QRasterBuffer;
108class QClipData;
109class QRasterPaintEngineState;
110
111template<typename F> class QRgbaFloat;
112typedef QRgbaFloat<float> QRgbaFloat32;
113
114typedef QT_FT_SpanFunc ProcessSpans;
115typedef void (*BitmapBlitFunc)(QRasterBuffer *rasterBuffer,
116 int x, int y, const QRgba64 &color,
117 const uchar *bitmap,
118 int mapWidth, int mapHeight, int mapStride);
119
120typedef void (*AlphamapBlitFunc)(QRasterBuffer *rasterBuffer,
121 int x, int y, const QRgba64 &color,
122 const uchar *bitmap,
123 int mapWidth, int mapHeight, int mapStride,
124 const QClipData *clip, bool useGammaCorrection);
125
126typedef void (*AlphaRGBBlitFunc)(QRasterBuffer *rasterBuffer,
127 int x, int y, const QRgba64 &color,
128 const uint *rgbmask,
129 int mapWidth, int mapHeight, int mapStride,
130 const QClipData *clip, bool useGammaCorrection);
131
132typedef void (*RectFillFunc)(QRasterBuffer *rasterBuffer,
133 int x, int y, int width, int height,
134 const QRgba64 &color);
135
136typedef void (*SrcOverBlendFunc)(uchar *destPixels, int dbpl,
137 const uchar *src, int spbl,
138 int w, int h,
139 int const_alpha);
140
141typedef void (*SrcOverScaleFunc)(uchar *destPixels, int dbpl,
142 const uchar *src, int spbl, int srch,
143 const QRectF &targetRect,
144 const QRectF &sourceRect,
145 const QRect &clipRect,
146 int const_alpha);
147
148typedef void (*SrcOverTransformFunc)(uchar *destPixels, int dbpl,
149 const uchar *src, int spbl,
150 const QRectF &targetRect,
151 const QRectF &sourceRect,
152 const QRect &clipRect,
153 const QTransform &targetRectTransform,
154 int const_alpha);
155
156struct DrawHelper {
157 ProcessSpans blendColor;
158 BitmapBlitFunc bitmapBlit;
159 AlphamapBlitFunc alphamapBlit;
160 AlphaRGBBlitFunc alphaRGBBlit;
161 RectFillFunc fillRect;
162};
163
164extern SrcOverBlendFunc qBlendFunctions[QImage::NImageFormats][QImage::NImageFormats];
165extern SrcOverScaleFunc qScaleFunctions[QImage::NImageFormats][QImage::NImageFormats];
166extern SrcOverTransformFunc qTransformFunctions[QImage::NImageFormats][QImage::NImageFormats];
167
168extern DrawHelper qDrawHelper[QImage::NImageFormats];
169
170struct quint24 {
171 quint24() = default;
172 quint24(uint value)
173 {
174 data[0] = uchar(value >> 16);
175 data[1] = uchar(value >> 8);
176 data[2] = uchar(value);
177 }
178 operator uint() const
179 {
180 return data[2] | (data[1] << 8) | (data[0] << 16);
181 }
182
183 uchar data[3];
184};
185
186void qBlendGradient(int count, const QSpan *spans, void *userData);
187void qBlendTexture(int count, const QSpan *spans, void *userData);
188#ifdef __SSE2__
189extern void (*qt_memfill64)(quint64 *dest, quint64 value, qsizetype count);
190extern void (*qt_memfill32)(quint32 *dest, quint32 value, qsizetype count);
191#else
192extern void qt_memfill64(quint64 *dest, quint64 value, qsizetype count);
193extern void qt_memfill32(quint32 *dest, quint32 value, qsizetype count);
194#endif
195extern void qt_memfill24(quint24 *dest, quint24 value, qsizetype count);
196extern void qt_memfill16(quint16 *dest, quint16 value, qsizetype count);
197
198typedef void (QT_FASTCALL *CompositionFunction)(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha);
199typedef void (QT_FASTCALL *CompositionFunction64)(QRgba64 *Q_DECL_RESTRICT dest, const QRgba64 *Q_DECL_RESTRICT src, int length, uint const_alpha);
200typedef void (QT_FASTCALL *CompositionFunctionFP)(QRgbaFloat32 *Q_DECL_RESTRICT dest, const QRgbaFloat32 *Q_DECL_RESTRICT src, int length, uint const_alpha);
201typedef void (QT_FASTCALL *CompositionFunctionSolid)(uint *dest, int length, uint color, uint const_alpha);
202typedef void (QT_FASTCALL *CompositionFunctionSolid64)(QRgba64 *dest, int length, QRgba64 color, uint const_alpha);
203typedef void (QT_FASTCALL *CompositionFunctionSolidFP)(QRgbaFloat32 *dest, int length, QRgbaFloat32 color, uint const_alpha);
204
205struct LinearGradientValues
206{
207 qreal dx;
208 qreal dy;
209 qreal l;
210 qreal off;
211};
212
213struct RadialGradientValues
214{
215 qreal dx;
216 qreal dy;
217 qreal dr;
218 qreal sqrfr;
219 qreal a;
220 qreal inv2a;
221 bool extended;
222};
223
224struct Operator;
225typedef uint* (QT_FASTCALL *DestFetchProc)(uint *buffer, QRasterBuffer *rasterBuffer, int x, int y, int length);
226typedef QRgba64* (QT_FASTCALL *DestFetchProc64)(QRgba64 *buffer, QRasterBuffer *rasterBuffer, int x, int y, int length);
227typedef QRgbaFloat32* (QT_FASTCALL *DestFetchProcFP)(QRgbaFloat32 *buffer, QRasterBuffer *rasterBuffer, int x, int y, int length);
228typedef void (QT_FASTCALL *DestStoreProc)(QRasterBuffer *rasterBuffer, int x, int y, const uint *buffer, int length);
229typedef void (QT_FASTCALL *DestStoreProc64)(QRasterBuffer *rasterBuffer, int x, int y, const QRgba64 *buffer, int length);
230typedef void (QT_FASTCALL *DestStoreProcFP)(QRasterBuffer *rasterBuffer, int x, int y, const QRgbaFloat32 *buffer, int length);
231typedef const uint* (QT_FASTCALL *SourceFetchProc)(uint *buffer, const Operator *o, const QSpanData *data, int y, int x, int length);
232typedef const QRgba64* (QT_FASTCALL *SourceFetchProc64)(QRgba64 *buffer, const Operator *o, const QSpanData *data, int y, int x, int length);
233typedef const QRgbaFloat32* (QT_FASTCALL *SourceFetchProcFP)(QRgbaFloat32 *buffer, const Operator *o, const QSpanData *data, int y, int x, int length);
234
235struct Operator
236{
237 QPainter::CompositionMode mode;
238 DestFetchProc destFetch;
239 DestStoreProc destStore;
240 SourceFetchProc srcFetch;
241 CompositionFunctionSolid funcSolid;
242 CompositionFunction func;
243
244 DestFetchProc64 destFetch64;
245 DestStoreProc64 destStore64;
246 SourceFetchProc64 srcFetch64;
247 CompositionFunctionSolid64 funcSolid64;
248 CompositionFunction64 func64;
249
250 DestFetchProcFP destFetchFP;
251 DestStoreProcFP destStoreFP;
252 SourceFetchProcFP srcFetchFP;
253 CompositionFunctionSolidFP funcSolidFP;
254 CompositionFunctionFP funcFP;
255
256 union {
257 LinearGradientValues linear;
258 RadialGradientValues radial;
259 };
260};
261
262class QRasterPaintEngine;
263
264struct QLinearGradientData
265{
266 struct {
267 qreal x;
268 qreal y;
269 } origin;
270 struct {
271 qreal x;
272 qreal y;
273 } end;
274};
275
276struct QRadialGradientData
277{
278 struct {
279 qreal x;
280 qreal y;
281 qreal radius;
282 } center;
283 struct {
284 qreal x;
285 qreal y;
286 qreal radius;
287 } focal;
288};
289
290struct QConicalGradientData
291{
292 struct {
293 qreal x;
294 qreal y;
295 } center;
296 qreal angle;
297};
298
299struct QGradientData
300{
301 QGradient::Spread spread;
302
303 union {
304 QLinearGradientData linear;
305 QRadialGradientData radial;
306 QConicalGradientData conical;
307 };
308
309#define GRADIENT_STOPTABLE_SIZE 1024
310#define GRADIENT_STOPTABLE_SIZE_SHIFT 10
311
312#if QT_CONFIG(raster_64bit) || QT_CONFIG(raster_fp)
313 const QRgba64 *colorTable64; //[GRADIENT_STOPTABLE_SIZE];
314#endif
315 const QRgb *colorTable32; //[GRADIENT_STOPTABLE_SIZE];
316
317 uint alphaColor : 1;
318};
319
320struct QTextureData
321{
322 const uchar *imageData;
323 const uchar *scanLine(int y) const { return imageData + y*bytesPerLine; }
324
325 int width;
326 int height;
327 // clip rect
328 int x1;
329 int y1;
330 int x2;
331 int y2;
332 qsizetype bytesPerLine;
333 QImage::Format format;
334 const QList<QRgb> *colorTable;
335 bool hasAlpha;
336 enum Type {
337 Plain,
338 Tiled,
339 Pattern
340 };
341 Type type;
342 int const_alpha;
343};
344
345struct QSpanData
346{
347 QSpanData() : tempImage(nullptr) {}
348 ~QSpanData() { delete tempImage; }
349
350 QRasterBuffer *rasterBuffer;
351 ProcessSpans blend;
352 ProcessSpans unclipped_blend;
353 BitmapBlitFunc bitmapBlit;
354 AlphamapBlitFunc alphamapBlit;
355 AlphaRGBBlitFunc alphaRGBBlit;
356 RectFillFunc fillRect;
357 qreal m11, m12, m13, m21, m22, m23, m33, dx, dy; // inverse xform matrix
358 const QClipData *clip;
359 enum Type {
360 None,
361 Solid,
362 LinearGradient,
363 RadialGradient,
364 ConicalGradient,
365 Texture
366 } type : 8;
367 signed int txop : 8;
368 uint fast_matrix : 1;
369 bool bilinear;
370 QImage *tempImage;
371 QRgba64 solidColor;
372 union {
373 QGradientData gradient;
374 QTextureData texture;
375 };
376 class Pinnable {
377 protected:
378 ~Pinnable() {}
379 }; // QSharedPointer<const void> is not supported
380 QSharedPointer<const Pinnable> cachedGradient;
381
382
383 void init(QRasterBuffer *rb, const QRasterPaintEngine *pe);
384 void setup(const QBrush &brush, int alpha, QPainter::CompositionMode compositionMode);
385 void setupMatrix(const QTransform &matrix, int bilinear);
386 void initTexture(const QImage *image, int alpha, QTextureData::Type = QTextureData::Plain, const QRect &sourceRect = QRect());
387 void adjustSpanMethods();
388};
389
390static inline uint qt_gradient_clamp(const QGradientData *data, int ipos)
391{
392 if (ipos < 0 || ipos >= GRADIENT_STOPTABLE_SIZE) {
393 if (data->spread == QGradient::RepeatSpread) {
394 ipos = ipos % GRADIENT_STOPTABLE_SIZE;
395 ipos = ipos < 0 ? GRADIENT_STOPTABLE_SIZE + ipos : ipos;
396 } else if (data->spread == QGradient::ReflectSpread) {
397 const int limit = GRADIENT_STOPTABLE_SIZE * 2;
398 ipos = ipos % limit;
399 ipos = ipos < 0 ? limit + ipos : ipos;
400 ipos = ipos >= GRADIENT_STOPTABLE_SIZE ? limit - 1 - ipos : ipos;
401 } else {
402 if (ipos < 0)
403 ipos = 0;
404 else if (ipos >= GRADIENT_STOPTABLE_SIZE)
405 ipos = GRADIENT_STOPTABLE_SIZE-1;
406 }
407 }
408
409 Q_ASSERT(ipos >= 0);
410 Q_ASSERT(ipos < GRADIENT_STOPTABLE_SIZE);
411
412 return ipos;
413}
414
415static inline uint qt_gradient_pixel(const QGradientData *data, qreal pos)
416{
417 int ipos = int(pos * (GRADIENT_STOPTABLE_SIZE - 1) + qreal(0.5));
418 return data->colorTable32[qt_gradient_clamp(data, ipos)];
419}
420
421#if QT_CONFIG(raster_64bit)
422static inline const QRgba64& qt_gradient_pixel64(const QGradientData *data, qreal pos)
423{
424 int ipos = int(pos * (GRADIENT_STOPTABLE_SIZE - 1) + qreal(0.5));
425 return data->colorTable64[qt_gradient_clamp(data, ipos)];
426}
427#endif
428
429static inline qreal qRadialDeterminant(qreal a, qreal b, qreal c)
430{
431 return (b * b) - (4 * a * c);
432}
433
434template <class RadialFetchFunc, typename BlendType> static
435const BlendType * QT_FASTCALL qt_fetch_radial_gradient_template(BlendType *buffer, const Operator *op,
436 const QSpanData *data, int y, int x, int length)
437{
438 // avoid division by zero
439 if (qFuzzyIsNull(op->radial.a)) {
440 RadialFetchFunc::memfill(buffer, RadialFetchFunc::null(), length);
441 return buffer;
442 }
443
444 const BlendType *b = buffer;
445 qreal rx = data->m21 * (y + qreal(0.5))
446 + data->dx + data->m11 * (x + qreal(0.5));
447 qreal ry = data->m22 * (y + qreal(0.5))
448 + data->dy + data->m12 * (x + qreal(0.5));
449 bool affine = !data->m13 && !data->m23;
450
451 BlendType *end = buffer + length;
452 if (affine) {
453 rx -= data->gradient.radial.focal.x;
454 ry -= data->gradient.radial.focal.y;
455
456 qreal inv_a = 1 / qreal(2 * op->radial.a);
457
458 const qreal delta_rx = data->m11;
459 const qreal delta_ry = data->m12;
460
461 qreal b = 2*(op->radial.dr*data->gradient.radial.focal.radius + rx * op->radial.dx + ry * op->radial.dy);
462 qreal delta_b = 2*(delta_rx * op->radial.dx + delta_ry * op->radial.dy);
463 const qreal b_delta_b = 2 * b * delta_b;
464 const qreal delta_b_delta_b = 2 * delta_b * delta_b;
465
466 const qreal bb = b * b;
467 const qreal delta_bb = delta_b * delta_b;
468
469 b *= inv_a;
470 delta_b *= inv_a;
471
472 const qreal rxrxryry = rx * rx + ry * ry;
473 const qreal delta_rxrxryry = delta_rx * delta_rx + delta_ry * delta_ry;
474 const qreal rx_plus_ry = 2*(rx * delta_rx + ry * delta_ry);
475 const qreal delta_rx_plus_ry = 2 * delta_rxrxryry;
476
477 inv_a *= inv_a;
478
479 qreal det = (bb - 4 * op->radial.a * (op->radial.sqrfr - rxrxryry)) * inv_a;
480 qreal delta_det = (b_delta_b + delta_bb + 4 * op->radial.a * (rx_plus_ry + delta_rxrxryry)) * inv_a;
481 const qreal delta_delta_det = (delta_b_delta_b + 4 * op->radial.a * delta_rx_plus_ry) * inv_a;
482
483 RadialFetchFunc::fetch(buffer, end, op, data, det, delta_det, delta_delta_det, b, delta_b);
484 } else {
485 qreal rw = data->m23 * (y + qreal(0.5))
486 + data->m33 + data->m13 * (x + qreal(0.5));
487
488 while (buffer < end) {
489 if (rw == 0) {
490 *buffer = RadialFetchFunc::null();
491 } else {
492 qreal invRw = 1 / rw;
493 qreal gx = rx * invRw - data->gradient.radial.focal.x;
494 qreal gy = ry * invRw - data->gradient.radial.focal.y;
495 qreal b = 2*(op->radial.dr*data->gradient.radial.focal.radius + gx*op->radial.dx + gy*op->radial.dy);
496 qreal det = qRadialDeterminant(op->radial.a, b, op->radial.sqrfr - (gx*gx + gy*gy));
497
498 BlendType result = RadialFetchFunc::null();
499 if (det >= 0) {
500 qreal detSqrt = qSqrt(det);
501
502 qreal s0 = (-b - detSqrt) * op->radial.inv2a;
503 qreal s1 = (-b + detSqrt) * op->radial.inv2a;
504
505 qreal s = qMax(s0, s1);
506
507 if (data->gradient.radial.focal.radius + op->radial.dr * s >= 0)
508 result = RadialFetchFunc::fetchSingle(data->gradient, s);
509 }
510
511 *buffer = result;
512 }
513
514 rx += data->m11;
515 ry += data->m12;
516 rw += data->m13;
517
518 ++buffer;
519 }
520 }
521
522 return b;
523}
524
525template <class Simd>
526class QRadialFetchSimd
527{
528public:
529 static uint null() { return 0; }
530 static uint fetchSingle(const QGradientData& gradient, qreal v)
531 {
532 return qt_gradient_pixel(&gradient, v);
533 }
534 static void memfill(uint *buffer, uint fill, int length)
535 {
536 qt_memfill32(buffer, fill, length);
537 }
538 static void fetch(uint *buffer, uint *end, const Operator *op, const QSpanData *data, qreal det,
539 qreal delta_det, qreal delta_delta_det, qreal b, qreal delta_b)
540 {
541 typename Simd::Vect_buffer_f det_vec;
542 typename Simd::Vect_buffer_f delta_det4_vec;
543 typename Simd::Vect_buffer_f b_vec;
544
545 for (int i = 0; i < 4; ++i) {
546 det_vec.f[i] = det;
547 delta_det4_vec.f[i] = 4 * delta_det;
548 b_vec.f[i] = b;
549
550 det += delta_det;
551 delta_det += delta_delta_det;
552 b += delta_b;
553 }
554
555 const typename Simd::Float32x4 v_delta_delta_det16 = Simd::v_dup(16 * delta_delta_det);
556 const typename Simd::Float32x4 v_delta_delta_det6 = Simd::v_dup(6 * delta_delta_det);
557 const typename Simd::Float32x4 v_delta_b4 = Simd::v_dup(4 * delta_b);
558
559 const typename Simd::Float32x4 v_r0 = Simd::v_dup(data->gradient.radial.focal.radius);
560 const typename Simd::Float32x4 v_dr = Simd::v_dup(op->radial.dr);
561
562#if defined(__ARM_NEON__)
563 // NEON doesn't have SIMD sqrt, but uses rsqrt instead that can't be taken of 0.
564 const typename Simd::Float32x4 v_min = Simd::v_dup(std::numeric_limits<float>::epsilon());
565#else
566 const typename Simd::Float32x4 v_min = Simd::v_dup(0.0f);
567#endif
568 const typename Simd::Float32x4 v_max = Simd::v_dup(float(GRADIENT_STOPTABLE_SIZE-1));
569 const typename Simd::Float32x4 v_half = Simd::v_dup(0.5f);
570
571 const typename Simd::Int32x4 v_repeat_mask = Simd::v_dup(~(uint(0xffffff) << GRADIENT_STOPTABLE_SIZE_SHIFT));
572 const typename Simd::Int32x4 v_reflect_mask = Simd::v_dup(~(uint(0xffffff) << (GRADIENT_STOPTABLE_SIZE_SHIFT+1)));
573
574 const typename Simd::Int32x4 v_reflect_limit = Simd::v_dup(2 * GRADIENT_STOPTABLE_SIZE - 1);
575
576 const int extended_mask = op->radial.extended ? 0x0 : ~0x0;
577
578#define FETCH_RADIAL_LOOP_PROLOGUE \
579 while (buffer < end) { \
580 typename Simd::Vect_buffer_i v_buffer_mask; \
581 v_buffer_mask.v = Simd::v_greaterOrEqual(det_vec.v, v_min); \
582 const typename Simd::Float32x4 v_index_local = Simd::v_sub(Simd::v_sqrt(Simd::v_max(v_min, det_vec.v)), b_vec.v); \
583 const typename Simd::Float32x4 v_index = Simd::v_add(Simd::v_mul(v_index_local, v_max), v_half); \
584 v_buffer_mask.v = Simd::v_and(v_buffer_mask.v, Simd::v_greaterOrEqual(Simd::v_add(v_r0, Simd::v_mul(v_dr, v_index_local)), v_min)); \
585 typename Simd::Vect_buffer_i index_vec;
586#define FETCH_RADIAL_LOOP_CLAMP_REPEAT \
587 index_vec.v = Simd::v_and(v_repeat_mask, Simd::v_toInt(v_index));
588#define FETCH_RADIAL_LOOP_CLAMP_REFLECT \
589 const typename Simd::Int32x4 v_index_i = Simd::v_and(v_reflect_mask, Simd::v_toInt(v_index)); \
590 const typename Simd::Int32x4 v_index_i_inv = Simd::v_sub(v_reflect_limit, v_index_i); \
591 index_vec.v = Simd::v_min_16(v_index_i, v_index_i_inv);
592#define FETCH_RADIAL_LOOP_CLAMP_PAD \
593 index_vec.v = Simd::v_toInt(Simd::v_min(v_max, Simd::v_max(v_min, v_index)));
594#define FETCH_RADIAL_LOOP_EPILOGUE \
595 det_vec.v = Simd::v_add(Simd::v_add(det_vec.v, delta_det4_vec.v), v_delta_delta_det6); \
596 delta_det4_vec.v = Simd::v_add(delta_det4_vec.v, v_delta_delta_det16); \
597 b_vec.v = Simd::v_add(b_vec.v, v_delta_b4); \
598 for (int i = 0; i < 4; ++i) \
599 *buffer++ = (extended_mask | v_buffer_mask.i[i]) & data->gradient.colorTable32[index_vec.i[i]]; \
600 }
601
602#define FETCH_RADIAL_LOOP(FETCH_RADIAL_LOOP_CLAMP) \
603 FETCH_RADIAL_LOOP_PROLOGUE \
604 FETCH_RADIAL_LOOP_CLAMP \
605 FETCH_RADIAL_LOOP_EPILOGUE
606
607 switch (data->gradient.spread) {
608 case QGradient::RepeatSpread:
609 FETCH_RADIAL_LOOP(FETCH_RADIAL_LOOP_CLAMP_REPEAT)
610 break;
611 case QGradient::ReflectSpread:
612 FETCH_RADIAL_LOOP(FETCH_RADIAL_LOOP_CLAMP_REFLECT)
613 break;
614 case QGradient::PadSpread:
615 FETCH_RADIAL_LOOP(FETCH_RADIAL_LOOP_CLAMP_PAD)
616 break;
617 default:
618 Q_UNREACHABLE();
619 }
620 }
621};
622
623static inline uint INTERPOLATE_PIXEL_255(uint x, uint a, uint y, uint b) {
624 uint t = (x & 0xff00ff) * a + (y & 0xff00ff) * b;
625 t = (t + ((t >> 8) & 0xff00ff) + 0x800080) >> 8;
626 t &= 0xff00ff;
627
628 x = ((x >> 8) & 0xff00ff) * a + ((y >> 8) & 0xff00ff) * b;
629 x = (x + ((x >> 8) & 0xff00ff) + 0x800080);
630 x &= 0xff00ff00;
631 x |= t;
632 return x;
633}
634
635#if Q_PROCESSOR_WORDSIZE == 8 // 64-bit versions
636
637static inline uint INTERPOLATE_PIXEL_256(uint x, uint a, uint y, uint b) {
638 quint64 t = (((quint64(x)) | ((quint64(x)) << 24)) & 0x00ff00ff00ff00ff) * a;
639 t += (((quint64(y)) | ((quint64(y)) << 24)) & 0x00ff00ff00ff00ff) * b;
640 t >>= 8;
641 t &= 0x00ff00ff00ff00ff;
642 return (uint(t)) | (uint(t >> 24));
643}
644
645static inline uint BYTE_MUL(uint x, uint a) {
646 quint64 t = (((quint64(x)) | ((quint64(x)) << 24)) & 0x00ff00ff00ff00ff) * a;
647 t = (t + ((t >> 8) & 0xff00ff00ff00ff) + 0x80008000800080) >> 8;
648 t &= 0x00ff00ff00ff00ff;
649 return (uint(t)) | (uint(t >> 24));
650}
651
652#else // 32-bit versions
653
654static inline uint INTERPOLATE_PIXEL_256(uint x, uint a, uint y, uint b) {
655 uint t = (x & 0xff00ff) * a + (y & 0xff00ff) * b;
656 t >>= 8;
657 t &= 0xff00ff;
658
659 x = ((x >> 8) & 0xff00ff) * a + ((y >> 8) & 0xff00ff) * b;
660 x &= 0xff00ff00;
661 x |= t;
662 return x;
663}
664
665static inline uint BYTE_MUL(uint x, uint a) {
666 uint t = (x & 0xff00ff) * a;
667 t = (t + ((t >> 8) & 0xff00ff) + 0x800080) >> 8;
668 t &= 0xff00ff;
669
670 x = ((x >> 8) & 0xff00ff) * a;
671 x = (x + ((x >> 8) & 0xff00ff) + 0x800080);
672 x &= 0xff00ff00;
673 x |= t;
674 return x;
675}
676#endif
677
678static inline void blend_pixel(quint32 &dst, const quint32 src)
679{
680 if (src >= 0xff000000)
681 dst = src;
682 else if (src != 0)
683 dst = src + BYTE_MUL(dst, qAlpha(~src));
684}
685
686static inline void blend_pixel(quint32 &dst, const quint32 src, const int const_alpha)
687{
688 if (const_alpha == 255)
689 return blend_pixel(dst, src);
690 if (src != 0) {
691 const quint32 s = BYTE_MUL(src, const_alpha);
692 dst = s + BYTE_MUL(dst, qAlpha(~s));
693 }
694}
695
696#if defined(__SSE2__)
697static inline uint Q_DECL_VECTORCALL interpolate_4_pixels_sse2(__m128i vt, __m128i vb, uint distx, uint disty)
698{
699 // First interpolate top and bottom pixels in parallel.
700 vt = _mm_unpacklo_epi8(vt, _mm_setzero_si128());
701 vb = _mm_unpacklo_epi8(vb, _mm_setzero_si128());
702 vt = _mm_mullo_epi16(vt, _mm_set1_epi16(256 - disty));
703 vb = _mm_mullo_epi16(vb, _mm_set1_epi16(disty));
704 __m128i vlr = _mm_add_epi16(vt, vb);
705 vlr = _mm_srli_epi16(vlr, 8);
706 // vlr now contains the result of the first two interpolate calls vlr = unpacked((xright << 64) | xleft)
707
708 // Now the last interpolate between left and right..
709 const __m128i vidistx = _mm_shufflelo_epi16(_mm_cvtsi32_si128(256 - distx), _MM_SHUFFLE(0, 0, 0, 0));
710 const __m128i vdistx = _mm_shufflelo_epi16(_mm_cvtsi32_si128(distx), _MM_SHUFFLE(0, 0, 0, 0));
711 const __m128i vmulx = _mm_unpacklo_epi16(vidistx, vdistx);
712 vlr = _mm_unpacklo_epi16(vlr, _mm_srli_si128(vlr, 8));
713 // vlr now contains the colors of left and right interleaved { la, ra, lr, rr, lg, rg, lb, rb }
714 vlr = _mm_madd_epi16(vlr, vmulx); // Multiply and horizontal add.
715 vlr = _mm_srli_epi32(vlr, 8);
716 vlr = _mm_packs_epi32(vlr, vlr);
717 vlr = _mm_packus_epi16(vlr, vlr);
718 return _mm_cvtsi128_si32(vlr);
719}
720
721static inline uint interpolate_4_pixels(uint tl, uint tr, uint bl, uint br, uint distx, uint disty)
722{
723 __m128i vt = _mm_unpacklo_epi32(_mm_cvtsi32_si128(tl), _mm_cvtsi32_si128(tr));
724 __m128i vb = _mm_unpacklo_epi32(_mm_cvtsi32_si128(bl), _mm_cvtsi32_si128(br));
725 return interpolate_4_pixels_sse2(vt, vb, distx, disty);
726}
727
728static inline uint interpolate_4_pixels(const uint t[], const uint b[], uint distx, uint disty)
729{
730 __m128i vt = _mm_loadl_epi64((const __m128i*)t);
731 __m128i vb = _mm_loadl_epi64((const __m128i*)b);
732 return interpolate_4_pixels_sse2(vt, vb, distx, disty);
733}
734
735static constexpr inline bool hasFastInterpolate4() { return true; }
736
737#elif defined(__ARM_NEON__)
738static inline uint interpolate_4_pixels_neon(uint32x2_t vt32, uint32x2_t vb32, uint distx, uint disty)
739{
740 uint16x8_t vt16 = vmovl_u8(vreinterpret_u8_u32(vt32));
741 uint16x8_t vb16 = vmovl_u8(vreinterpret_u8_u32(vb32));
742 vt16 = vmulq_n_u16(vt16, 256 - disty);
743 vt16 = vmlaq_n_u16(vt16, vb16, disty);
744 vt16 = vshrq_n_u16(vt16, 8);
745 uint16x4_t vl16 = vget_low_u16(vt16);
746 uint16x4_t vr16 = vget_high_u16(vt16);
747 vl16 = vmul_n_u16(vl16, 256 - distx);
748 vl16 = vmla_n_u16(vl16, vr16, distx);
749 vl16 = vshr_n_u16(vl16, 8);
750 uint8x8_t vr = vmovn_u16(vcombine_u16(vl16, vl16));
751 return vget_lane_u32(vreinterpret_u32_u8(vr), 0);
752}
753
754static inline uint interpolate_4_pixels(uint tl, uint tr, uint bl, uint br, uint distx, uint disty)
755{
756 uint32x2_t vt32 = vmov_n_u32(tl);
757 uint32x2_t vb32 = vmov_n_u32(bl);
758 vt32 = vset_lane_u32(tr, vt32, 1);
759 vb32 = vset_lane_u32(br, vb32, 1);
760 return interpolate_4_pixels_neon(vt32, vb32, distx, disty);
761}
762
763static inline uint interpolate_4_pixels(const uint t[], const uint b[], uint distx, uint disty)
764{
765 uint32x2_t vt32 = vld1_u32(t);
766 uint32x2_t vb32 = vld1_u32(b);
767 return interpolate_4_pixels_neon(vt32, vb32, distx, disty);
768}
769
770static constexpr inline bool hasFastInterpolate4() { return true; }
771
772#else
773static inline uint interpolate_4_pixels(uint tl, uint tr, uint bl, uint br, uint distx, uint disty)
774{
775 uint idistx = 256 - distx;
776 uint idisty = 256 - disty;
777 uint xtop = INTERPOLATE_PIXEL_256(tl, idistx, tr, distx);
778 uint xbot = INTERPOLATE_PIXEL_256(bl, idistx, br, distx);
779 return INTERPOLATE_PIXEL_256(xtop, idisty, xbot, disty);
780}
781
782static inline uint interpolate_4_pixels(const uint t[], const uint b[], uint distx, uint disty)
783{
784 return interpolate_4_pixels(t[0], t[1], b[0], b[1], distx, disty);
785}
786
787static constexpr inline bool hasFastInterpolate4() { return false; }
788
789#endif
790
791static inline QRgba64 multiplyAlpha256(QRgba64 rgba64, uint alpha256)
792{
793 return QRgba64::fromRgba64((rgba64.red() * alpha256) >> 8,
794 (rgba64.green() * alpha256) >> 8,
795 (rgba64.blue() * alpha256) >> 8,
796 (rgba64.alpha() * alpha256) >> 8);
797}
798static inline QRgba64 interpolate256(QRgba64 x, uint alpha1, QRgba64 y, uint alpha2)
799{
800 return QRgba64::fromRgba64(multiplyAlpha256(x, alpha1) + multiplyAlpha256(y, alpha2));
801}
802
803#ifdef __SSE2__
804static inline QRgba64 interpolate_4_pixels_rgb64(const QRgba64 t[], const QRgba64 b[], uint distx, uint disty)
805{
806 __m128i vt = _mm_loadu_si128((const __m128i*)t);
807 if (disty) {
808 __m128i vb = _mm_loadu_si128((const __m128i*)b);
809 vt = _mm_mulhi_epu16(vt, _mm_set1_epi16(0x10000 - disty));
810 vb = _mm_mulhi_epu16(vb, _mm_set1_epi16(disty));
811 vt = _mm_add_epi16(vt, vb);
812 }
813 if (distx) {
814 const __m128i vdistx = _mm_shufflelo_epi16(_mm_cvtsi32_si128(distx), _MM_SHUFFLE(0, 0, 0, 0));
815 const __m128i vidistx = _mm_shufflelo_epi16(_mm_cvtsi32_si128(0x10000 - distx), _MM_SHUFFLE(0, 0, 0, 0));
816 vt = _mm_mulhi_epu16(vt, _mm_unpacklo_epi64(vidistx, vdistx));
817 vt = _mm_add_epi16(vt, _mm_srli_si128(vt, 8));
818 }
819#ifdef Q_PROCESSOR_X86_64
820 return QRgba64::fromRgba64(_mm_cvtsi128_si64(vt));
821#else
822 QRgba64 out;
823 _mm_storel_epi64((__m128i*)&out, vt);
824 return out;
825#endif // Q_PROCESSOR_X86_64
826}
827#elif defined(__ARM_NEON__)
828static inline QRgba64 interpolate_4_pixels_rgb64(const QRgba64 t[], const QRgba64 b[], uint distx, uint disty)
829{
830 uint64x1x2_t vt = vld2_u64(reinterpret_cast<const uint64_t *>(t));
831 if (disty) {
832 uint64x1x2_t vb = vld2_u64(reinterpret_cast<const uint64_t *>(b));
833 uint32x4_t vt0 = vmull_n_u16(vreinterpret_u16_u64(vt.val[0]), 0x10000 - disty);
834 uint32x4_t vt1 = vmull_n_u16(vreinterpret_u16_u64(vt.val[1]), 0x10000 - disty);
835 vt0 = vmlal_n_u16(vt0, vreinterpret_u16_u64(vb.val[0]), disty);
836 vt1 = vmlal_n_u16(vt1, vreinterpret_u16_u64(vb.val[1]), disty);
837 vt.val[0] = vreinterpret_u64_u16(vshrn_n_u32(vt0, 16));
838 vt.val[1] = vreinterpret_u64_u16(vshrn_n_u32(vt1, 16));
839 }
840 if (distx) {
841 uint32x4_t vt0 = vmull_n_u16(vreinterpret_u16_u64(vt.val[0]), 0x10000 - distx);
842 vt0 = vmlal_n_u16(vt0, vreinterpret_u16_u64(vt.val[1]), distx);
843 vt.val[0] = vreinterpret_u64_u16(vshrn_n_u32(vt0, 16));
844 }
845 QRgba64 out;
846 vst1_u64(reinterpret_cast<uint64_t *>(&out), vt.val[0]);
847 return out;
848}
849#else
850static inline QRgba64 interpolate_4_pixels_rgb64(const QRgba64 t[], const QRgba64 b[], uint distx, uint disty)
851{
852 const uint dx = distx>>8;
853 const uint dy = disty>>8;
854 const uint idx = 256 - dx;
855 const uint idy = 256 - dy;
856 QRgba64 xtop = interpolate256(t[0], idx, t[1], dx);
857 QRgba64 xbot = interpolate256(b[0], idx, b[1], dx);
858 return interpolate256(xtop, idy, xbot, dy);
859}
860#endif // __SSE2__
861
862#if QT_CONFIG(raster_fp)
863static inline QRgbaFloat32 multiplyAlpha_rgba32f(QRgbaFloat32 c, float a)
864{
865 return QRgbaFloat32 { c.r * a, c.g * a, c.b * a, c.a * a };
866}
867
868static inline QRgbaFloat32 interpolate_rgba32f(QRgbaFloat32 x, float alpha1, QRgbaFloat32 y, float alpha2)
869{
870 x = multiplyAlpha_rgba32f(x, alpha1);
871 y = multiplyAlpha_rgba32f(y, alpha2);
872 return QRgbaFloat32 { x.r + y.r, x.g + y.g, x.b + y.b, x.a + y.a };
873}
874#ifdef __SSE2__
875static inline __m128 Q_DECL_VECTORCALL interpolate_rgba32f(__m128 x, __m128 alpha1, __m128 y, __m128 alpha2)
876{
877 return _mm_add_ps(_mm_mul_ps(x, alpha1), _mm_mul_ps(y, alpha2));
878}
879#endif
880
881static inline QRgbaFloat32 interpolate_4_pixels_rgba32f(const QRgbaFloat32 t[], const QRgbaFloat32 b[], uint distx, uint disty)
882{
883 constexpr float f = 1.0f / 65536.0f;
884 const float dx = distx * f;
885 const float dy = disty * f;
886 const float idx = 1.0f - dx;
887 const float idy = 1.0f - dy;
888#ifdef __SSE2__
889 const __m128 vtl = _mm_load_ps((const float *)&t[0]);
890 const __m128 vtr = _mm_load_ps((const float *)&t[1]);
891 const __m128 vbl = _mm_load_ps((const float *)&b[0]);
892 const __m128 vbr = _mm_load_ps((const float *)&b[1]);
893
894 const __m128 vdx = _mm_set1_ps(dx);
895 const __m128 vidx = _mm_set1_ps(idx);
896 __m128 vt = interpolate_rgba32f(vtl, vidx, vtr, vdx);
897 __m128 vb = interpolate_rgba32f(vbl, vidx, vbr, vdx);
898 const __m128 vdy = _mm_set1_ps(dy);
899 const __m128 vidy = _mm_set1_ps(idy);
900 vt = interpolate_rgba32f(vt, vidy, vb, vdy);
901 QRgbaFloat32 res;
902 _mm_store_ps((float*)&res, vt);
903 return res;
904#else
905 QRgbaFloat32 xtop = interpolate_rgba32f(t[0], idx, t[1], dx);
906 QRgbaFloat32 xbot = interpolate_rgba32f(b[0], idx, b[1], dx);
907 xtop = interpolate_rgba32f(xtop, idy, xbot, dy);
908 return xtop;
909#endif
910}
911#endif // QT_CONFIG(raster_fp)
912
913static inline uint BYTE_MUL_RGB16(uint x, uint a) {
914 a += 1;
915 uint t = (((x & 0x07e0)*a) >> 8) & 0x07e0;
916 t |= (((x & 0xf81f)*(a>>2)) >> 6) & 0xf81f;
917 return t;
918}
919
920static inline uint BYTE_MUL_RGB16_32(uint x, uint a) {
921 uint t = (((x & 0xf81f07e0) >> 5)*a) & 0xf81f07e0;
922 t |= (((x & 0x07e0f81f)*a) >> 5) & 0x07e0f81f;
923 return t;
924}
925
926// qt_div_255 is a fast rounded division by 255 using an approximation that is accurate for all positive 16-bit integers
927static constexpr inline int qt_div_255(int x) { return (x + (x>>8) + 0x80) >> 8; }
928static constexpr inline uint qt_div_257_floor(uint x) { return (x - (x >> 8)) >> 8; }
929static constexpr inline uint qt_div_257(uint x) { return qt_div_257_floor(x + 128); }
930static constexpr inline uint qt_div_65535(uint x) { return (x + (x>>16) + 0x8000U) >> 16; }
931
932template <class T> inline void qt_memfill_template(T *dest, T color, qsizetype count)
933{
934 if (!count)
935 return;
936
937 qsizetype n = (count + 7) / 8;
938 switch (count & 0x07)
939 {
940 case 0: do { *dest++ = color; Q_FALLTHROUGH();
941 case 7: *dest++ = color; Q_FALLTHROUGH();
942 case 6: *dest++ = color; Q_FALLTHROUGH();
943 case 5: *dest++ = color; Q_FALLTHROUGH();
944 case 4: *dest++ = color; Q_FALLTHROUGH();
945 case 3: *dest++ = color; Q_FALLTHROUGH();
946 case 2: *dest++ = color; Q_FALLTHROUGH();
947 case 1: *dest++ = color;
948 } while (--n > 0);
949 }
950}
951
952template <class T> inline void qt_memfill(T *dest, T value, qsizetype count)
953{
954 qt_memfill_template(dest, value, count);
955}
956
957template<> inline void qt_memfill(quint64 *dest, quint64 color, qsizetype count)
958{
959 qt_memfill64(dest, color, count);
960}
961
962template<> inline void qt_memfill(quint32 *dest, quint32 color, qsizetype count)
963{
964 qt_memfill32(dest, color, count);
965}
966
967template<> inline void qt_memfill(quint24 *dest, quint24 color, qsizetype count)
968{
969 qt_memfill24(dest, color, count);
970}
971
972template<> inline void qt_memfill(quint16 *dest, quint16 color, qsizetype count)
973{
974 qt_memfill16(dest, color, count);
975}
976
977template<> inline void qt_memfill(quint8 *dest, quint8 color, qsizetype count)
978{
979 memset(dest, color, count);
980}
981
982template <class T> static
983inline void qt_rectfill(T *dest, T value,
984 int x, int y, int width, int height, qsizetype stride)
985{
986 char *d = reinterpret_cast<char*>(dest + x) + y * stride;
987 if (uint(stride) == (width * sizeof(T))) {
988 qt_memfill(reinterpret_cast<T*>(d), value, qsizetype(width) * height);
989 } else {
990 for (int j = 0; j < height; ++j) {
991 dest = reinterpret_cast<T*>(d);
992 qt_memfill(dest, value, width);
993 d += stride;
994 }
995 }
996}
997
998inline ushort qConvertRgb32To16(uint c)
999{
1000 return (((c) >> 3) & 0x001f)
1001 | (((c) >> 5) & 0x07e0)
1002 | (((c) >> 8) & 0xf800);
1003}
1004
1005inline QRgb qConvertRgb16To32(uint c)
1006{
1007 return 0xff000000
1008 | ((((c) << 3) & 0xf8) | (((c) >> 2) & 0x7))
1009 | ((((c) << 5) & 0xfc00) | (((c) >> 1) & 0x300))
1010 | ((((c) << 8) & 0xf80000) | (((c) << 3) & 0x70000));
1011}
1012
1013const uint qt_bayer_matrix[16][16] = {
1014 { 0x1, 0xc0, 0x30, 0xf0, 0xc, 0xcc, 0x3c, 0xfc,
1015 0x3, 0xc3, 0x33, 0xf3, 0xf, 0xcf, 0x3f, 0xff},
1016 { 0x80, 0x40, 0xb0, 0x70, 0x8c, 0x4c, 0xbc, 0x7c,
1017 0x83, 0x43, 0xb3, 0x73, 0x8f, 0x4f, 0xbf, 0x7f},
1018 { 0x20, 0xe0, 0x10, 0xd0, 0x2c, 0xec, 0x1c, 0xdc,
1019 0x23, 0xe3, 0x13, 0xd3, 0x2f, 0xef, 0x1f, 0xdf},
1020 { 0xa0, 0x60, 0x90, 0x50, 0xac, 0x6c, 0x9c, 0x5c,
1021 0xa3, 0x63, 0x93, 0x53, 0xaf, 0x6f, 0x9f, 0x5f},
1022 { 0x8, 0xc8, 0x38, 0xf8, 0x4, 0xc4, 0x34, 0xf4,
1023 0xb, 0xcb, 0x3b, 0xfb, 0x7, 0xc7, 0x37, 0xf7},
1024 { 0x88, 0x48, 0xb8, 0x78, 0x84, 0x44, 0xb4, 0x74,
1025 0x8b, 0x4b, 0xbb, 0x7b, 0x87, 0x47, 0xb7, 0x77},
1026 { 0x28, 0xe8, 0x18, 0xd8, 0x24, 0xe4, 0x14, 0xd4,
1027 0x2b, 0xeb, 0x1b, 0xdb, 0x27, 0xe7, 0x17, 0xd7},
1028 { 0xa8, 0x68, 0x98, 0x58, 0xa4, 0x64, 0x94, 0x54,
1029 0xab, 0x6b, 0x9b, 0x5b, 0xa7, 0x67, 0x97, 0x57},
1030 { 0x2, 0xc2, 0x32, 0xf2, 0xe, 0xce, 0x3e, 0xfe,
1031 0x1, 0xc1, 0x31, 0xf1, 0xd, 0xcd, 0x3d, 0xfd},
1032 { 0x82, 0x42, 0xb2, 0x72, 0x8e, 0x4e, 0xbe, 0x7e,
1033 0x81, 0x41, 0xb1, 0x71, 0x8d, 0x4d, 0xbd, 0x7d},
1034 { 0x22, 0xe2, 0x12, 0xd2, 0x2e, 0xee, 0x1e, 0xde,
1035 0x21, 0xe1, 0x11, 0xd1, 0x2d, 0xed, 0x1d, 0xdd},
1036 { 0xa2, 0x62, 0x92, 0x52, 0xae, 0x6e, 0x9e, 0x5e,
1037 0xa1, 0x61, 0x91, 0x51, 0xad, 0x6d, 0x9d, 0x5d},
1038 { 0xa, 0xca, 0x3a, 0xfa, 0x6, 0xc6, 0x36, 0xf6,
1039 0x9, 0xc9, 0x39, 0xf9, 0x5, 0xc5, 0x35, 0xf5},
1040 { 0x8a, 0x4a, 0xba, 0x7a, 0x86, 0x46, 0xb6, 0x76,
1041 0x89, 0x49, 0xb9, 0x79, 0x85, 0x45, 0xb5, 0x75},
1042 { 0x2a, 0xea, 0x1a, 0xda, 0x26, 0xe6, 0x16, 0xd6,
1043 0x29, 0xe9, 0x19, 0xd9, 0x25, 0xe5, 0x15, 0xd5},
1044 { 0xaa, 0x6a, 0x9a, 0x5a, 0xa6, 0x66, 0x96, 0x56,
1045 0xa9, 0x69, 0x99, 0x59, 0xa5, 0x65, 0x95, 0x55}
1046};
1047
1048#define ARGB_COMBINE_ALPHA(argb, alpha) \
1049 ((((argb >> 24) * alpha) >> 8) << 24) | (argb & 0x00ffffff)
1050
1051
1052#if Q_PROCESSOR_WORDSIZE == 8 // 64-bit versions
1053#define AMIX(mask) (qMin(((quint64(s)&mask) + (quint64(d)&mask)), quint64(mask)))
1054#define MIX(mask) (qMin(((quint64(s)&mask) + (quint64(d)&mask)), quint64(mask)))
1055#else // 32 bits
1056// The mask for alpha can overflow over 32 bits
1057#define AMIX(mask) quint32(qMin(((quint64(s)&mask) + (quint64(d)&mask)), quint64(mask)))
1058#define MIX(mask) (qMin(((quint32(s)&mask) + (quint32(d)&mask)), quint32(mask)))
1059#endif
1060
1061inline uint comp_func_Plus_one_pixel_const_alpha(uint d, const uint s, const uint const_alpha, const uint one_minus_const_alpha)
1062{
1063 const uint result = uint(AMIX(AMASK) | MIX(RMASK) | MIX(GMASK) | MIX(BMASK));
1064 return INTERPOLATE_PIXEL_255(result, const_alpha, d, one_minus_const_alpha);
1065}
1066
1067inline uint comp_func_Plus_one_pixel(uint d, const uint s)
1068{
1069 const uint result = uint(AMIX(AMASK) | MIX(RMASK) | MIX(GMASK) | MIX(BMASK));
1070 return result;
1071}
1072
1073#undef MIX
1074#undef AMIX
1075
1076// must be multiple of 4 for easier SIMD implementations
1077static constexpr int BufferSize = 2048;
1078
1079// A buffer of intermediate results used by simple bilinear scaling.
1080struct IntermediateBuffer
1081{
1082 // The idea is first to do the interpolation between the row s1 and the row s2
1083 // into this intermediate buffer, then later interpolate between two pixel of this buffer.
1084 //
1085 // buffer_rb is a buffer of red-blue component of the pixel, in the form 0x00RR00BB
1086 // buffer_ag is the alpha-green component of the pixel, in the form 0x00AA00GG
1087 // +1 for the last pixel to interpolate with, and +1 for rounding errors.
1088 quint32 buffer_rb[BufferSize+2];
1089 quint32 buffer_ag[BufferSize+2];
1090};
1091
1092QT_END_NAMESPACE
1093
1094#endif // QDRAWHELPER_P_H
1095

source code of qtbase/src/gui/painting/qdrawhelper_p.h