| 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 |
| 62 | #define |
| 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 | |
| 70 | QT_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 | |
| 87 | static const uint AMASK = 0xff000000; |
| 88 | static const uint RMASK = 0x00ff0000; |
| 89 | static const uint GMASK = 0x0000ff00; |
| 90 | static const uint BMASK = 0x000000ff; |
| 91 | |
| 92 | /******************************************************************************* |
| 93 | * QSpan |
| 94 | * |
| 95 | * duplicate definition of FT_Span |
| 96 | */ |
| 97 | typedef QT_FT_Span QSpan; |
| 98 | |
| 99 | struct QSolidData; |
| 100 | struct QTextureData; |
| 101 | struct QGradientData; |
| 102 | struct QLinearGradientData; |
| 103 | struct QRadialGradientData; |
| 104 | struct QConicalGradientData; |
| 105 | struct QSpanData; |
| 106 | class QGradient; |
| 107 | class QRasterBuffer; |
| 108 | class QClipData; |
| 109 | class QRasterPaintEngineState; |
| 110 | |
| 111 | typedef QT_FT_SpanFunc ProcessSpans; |
| 112 | typedef void (*BitmapBlitFunc)(QRasterBuffer *rasterBuffer, |
| 113 | int x, int y, const QRgba64 &color, |
| 114 | const uchar *bitmap, |
| 115 | int mapWidth, int mapHeight, int mapStride); |
| 116 | |
| 117 | typedef void (*AlphamapBlitFunc)(QRasterBuffer *rasterBuffer, |
| 118 | int x, int y, const QRgba64 &color, |
| 119 | const uchar *bitmap, |
| 120 | int mapWidth, int mapHeight, int mapStride, |
| 121 | const QClipData *clip, bool useGammaCorrection); |
| 122 | |
| 123 | typedef void (*AlphaRGBBlitFunc)(QRasterBuffer *rasterBuffer, |
| 124 | int x, int y, const QRgba64 &color, |
| 125 | const uint *rgbmask, |
| 126 | int mapWidth, int mapHeight, int mapStride, |
| 127 | const QClipData *clip, bool useGammaCorrection); |
| 128 | |
| 129 | typedef void (*RectFillFunc)(QRasterBuffer *rasterBuffer, |
| 130 | int x, int y, int width, int height, |
| 131 | const QRgba64 &color); |
| 132 | |
| 133 | typedef void (*SrcOverBlendFunc)(uchar *destPixels, int dbpl, |
| 134 | const uchar *src, int spbl, |
| 135 | int w, int h, |
| 136 | int const_alpha); |
| 137 | |
| 138 | typedef void (*SrcOverScaleFunc)(uchar *destPixels, int dbpl, |
| 139 | const uchar *src, int spbl, int srch, |
| 140 | const QRectF &targetRect, |
| 141 | const QRectF &sourceRect, |
| 142 | const QRect &clipRect, |
| 143 | int const_alpha); |
| 144 | |
| 145 | typedef void (*SrcOverTransformFunc)(uchar *destPixels, int dbpl, |
| 146 | const uchar *src, int spbl, |
| 147 | const QRectF &targetRect, |
| 148 | const QRectF &sourceRect, |
| 149 | const QRect &clipRect, |
| 150 | const QTransform &targetRectTransform, |
| 151 | int const_alpha); |
| 152 | |
| 153 | struct DrawHelper { |
| 154 | ProcessSpans blendColor; |
| 155 | BitmapBlitFunc bitmapBlit; |
| 156 | AlphamapBlitFunc alphamapBlit; |
| 157 | AlphaRGBBlitFunc alphaRGBBlit; |
| 158 | RectFillFunc fillRect; |
| 159 | }; |
| 160 | |
| 161 | extern SrcOverBlendFunc qBlendFunctions[QImage::NImageFormats][QImage::NImageFormats]; |
| 162 | extern SrcOverScaleFunc qScaleFunctions[QImage::NImageFormats][QImage::NImageFormats]; |
| 163 | extern SrcOverTransformFunc qTransformFunctions[QImage::NImageFormats][QImage::NImageFormats]; |
| 164 | |
| 165 | extern DrawHelper qDrawHelper[QImage::NImageFormats]; |
| 166 | |
| 167 | struct quint24 { |
| 168 | quint24() = default; |
| 169 | quint24(uint value) |
| 170 | { |
| 171 | data[0] = uchar(value >> 16); |
| 172 | data[1] = uchar(value >> 8); |
| 173 | data[2] = uchar(value); |
| 174 | } |
| 175 | operator uint() const |
| 176 | { |
| 177 | return data[2] | (data[1] << 8) | (data[0] << 16); |
| 178 | } |
| 179 | |
| 180 | uchar data[3]; |
| 181 | }; |
| 182 | |
| 183 | void qBlendGradient(int count, const QSpan *spans, void *userData); |
| 184 | void qBlendTexture(int count, const QSpan *spans, void *userData); |
| 185 | #ifdef __SSE2__ |
| 186 | extern void (*qt_memfill64)(quint64 *dest, quint64 value, qsizetype count); |
| 187 | extern void (*qt_memfill32)(quint32 *dest, quint32 value, qsizetype count); |
| 188 | #else |
| 189 | extern void qt_memfill64(quint64 *dest, quint64 value, qsizetype count); |
| 190 | extern void qt_memfill32(quint32 *dest, quint32 value, qsizetype count); |
| 191 | #endif |
| 192 | extern void qt_memfill24(quint24 *dest, quint24 value, qsizetype count); |
| 193 | extern void qt_memfill16(quint16 *dest, quint16 value, qsizetype count); |
| 194 | |
| 195 | typedef void (QT_FASTCALL *CompositionFunction)(uint *Q_DECL_RESTRICT dest, const uint *Q_DECL_RESTRICT src, int length, uint const_alpha); |
| 196 | typedef void (QT_FASTCALL *CompositionFunction64)(QRgba64 *Q_DECL_RESTRICT dest, const QRgba64 *Q_DECL_RESTRICT src, int length, uint const_alpha); |
| 197 | typedef void (QT_FASTCALL *CompositionFunctionSolid)(uint *dest, int length, uint color, uint const_alpha); |
| 198 | typedef void (QT_FASTCALL *CompositionFunctionSolid64)(QRgba64 *dest, int length, QRgba64 color, uint const_alpha); |
| 199 | |
| 200 | struct LinearGradientValues |
| 201 | { |
| 202 | qreal dx; |
| 203 | qreal dy; |
| 204 | qreal l; |
| 205 | qreal off; |
| 206 | }; |
| 207 | |
| 208 | struct RadialGradientValues |
| 209 | { |
| 210 | qreal dx; |
| 211 | qreal dy; |
| 212 | qreal dr; |
| 213 | qreal sqrfr; |
| 214 | qreal a; |
| 215 | qreal inv2a; |
| 216 | bool extended; |
| 217 | }; |
| 218 | |
| 219 | struct Operator; |
| 220 | typedef uint* (QT_FASTCALL *DestFetchProc)(uint *buffer, QRasterBuffer *rasterBuffer, int x, int y, int length); |
| 221 | typedef QRgba64* (QT_FASTCALL *DestFetchProc64)(QRgba64 *buffer, QRasterBuffer *rasterBuffer, int x, int y, int length); |
| 222 | typedef void (QT_FASTCALL *DestStoreProc)(QRasterBuffer *rasterBuffer, int x, int y, const uint *buffer, int length); |
| 223 | typedef void (QT_FASTCALL *DestStoreProc64)(QRasterBuffer *rasterBuffer, int x, int y, const QRgba64 *buffer, int length); |
| 224 | typedef const uint* (QT_FASTCALL *SourceFetchProc)(uint *buffer, const Operator *o, const QSpanData *data, int y, int x, int length); |
| 225 | typedef const QRgba64* (QT_FASTCALL *SourceFetchProc64)(QRgba64 *buffer, const Operator *o, const QSpanData *data, int y, int x, int length); |
| 226 | |
| 227 | struct Operator |
| 228 | { |
| 229 | QPainter::CompositionMode mode; |
| 230 | DestFetchProc destFetch; |
| 231 | DestStoreProc destStore; |
| 232 | SourceFetchProc srcFetch; |
| 233 | CompositionFunctionSolid funcSolid; |
| 234 | CompositionFunction func; |
| 235 | |
| 236 | DestFetchProc64 destFetch64; |
| 237 | DestStoreProc64 destStore64; |
| 238 | SourceFetchProc64 srcFetch64; |
| 239 | CompositionFunctionSolid64 funcSolid64; |
| 240 | CompositionFunction64 func64; |
| 241 | |
| 242 | union { |
| 243 | LinearGradientValues linear; |
| 244 | RadialGradientValues radial; |
| 245 | }; |
| 246 | }; |
| 247 | |
| 248 | class QRasterPaintEngine; |
| 249 | |
| 250 | struct QLinearGradientData |
| 251 | { |
| 252 | struct { |
| 253 | qreal x; |
| 254 | qreal y; |
| 255 | } origin; |
| 256 | struct { |
| 257 | qreal x; |
| 258 | qreal y; |
| 259 | } end; |
| 260 | }; |
| 261 | |
| 262 | struct QRadialGradientData |
| 263 | { |
| 264 | struct { |
| 265 | qreal x; |
| 266 | qreal y; |
| 267 | qreal radius; |
| 268 | } center; |
| 269 | struct { |
| 270 | qreal x; |
| 271 | qreal y; |
| 272 | qreal radius; |
| 273 | } focal; |
| 274 | }; |
| 275 | |
| 276 | struct QConicalGradientData |
| 277 | { |
| 278 | struct { |
| 279 | qreal x; |
| 280 | qreal y; |
| 281 | } center; |
| 282 | qreal angle; |
| 283 | }; |
| 284 | |
| 285 | struct QGradientData |
| 286 | { |
| 287 | QGradient::Spread spread; |
| 288 | |
| 289 | union { |
| 290 | QLinearGradientData linear; |
| 291 | QRadialGradientData radial; |
| 292 | QConicalGradientData conical; |
| 293 | }; |
| 294 | |
| 295 | #define GRADIENT_STOPTABLE_SIZE 1024 |
| 296 | #define GRADIENT_STOPTABLE_SIZE_SHIFT 10 |
| 297 | |
| 298 | #if QT_CONFIG(raster_64bit) |
| 299 | const QRgba64 *colorTable64; //[GRADIENT_STOPTABLE_SIZE]; |
| 300 | #endif |
| 301 | const QRgb *colorTable32; //[GRADIENT_STOPTABLE_SIZE]; |
| 302 | |
| 303 | uint alphaColor : 1; |
| 304 | }; |
| 305 | |
| 306 | struct QTextureData |
| 307 | { |
| 308 | const uchar *imageData; |
| 309 | const uchar *scanLine(int y) const { return imageData + y*bytesPerLine; } |
| 310 | |
| 311 | int width; |
| 312 | int height; |
| 313 | // clip rect |
| 314 | int x1; |
| 315 | int y1; |
| 316 | int x2; |
| 317 | int y2; |
| 318 | qsizetype bytesPerLine; |
| 319 | QImage::Format format; |
| 320 | const QList<QRgb> *colorTable; |
| 321 | bool hasAlpha; |
| 322 | enum Type { |
| 323 | Plain, |
| 324 | Tiled |
| 325 | }; |
| 326 | Type type; |
| 327 | int const_alpha; |
| 328 | }; |
| 329 | |
| 330 | struct QSpanData |
| 331 | { |
| 332 | QSpanData() : tempImage(nullptr) {} |
| 333 | ~QSpanData() { delete tempImage; } |
| 334 | |
| 335 | QRasterBuffer *rasterBuffer; |
| 336 | ProcessSpans blend; |
| 337 | ProcessSpans unclipped_blend; |
| 338 | BitmapBlitFunc bitmapBlit; |
| 339 | AlphamapBlitFunc alphamapBlit; |
| 340 | AlphaRGBBlitFunc alphaRGBBlit; |
| 341 | RectFillFunc fillRect; |
| 342 | qreal m11, m12, m13, m21, m22, m23, m33, dx, dy; // inverse xform matrix |
| 343 | const QClipData *clip; |
| 344 | enum Type { |
| 345 | None, |
| 346 | Solid, |
| 347 | LinearGradient, |
| 348 | RadialGradient, |
| 349 | ConicalGradient, |
| 350 | Texture |
| 351 | } type : 8; |
| 352 | signed int txop : 8; |
| 353 | uint fast_matrix : 1; |
| 354 | bool bilinear; |
| 355 | QImage *tempImage; |
| 356 | QRgba64 solidColor; |
| 357 | union { |
| 358 | QGradientData gradient; |
| 359 | QTextureData texture; |
| 360 | }; |
| 361 | class Pinnable { |
| 362 | protected: |
| 363 | ~Pinnable() {} |
| 364 | }; // QSharedPointer<const void> is not supported |
| 365 | QSharedPointer<const Pinnable> cachedGradient; |
| 366 | |
| 367 | |
| 368 | void init(QRasterBuffer *rb, const QRasterPaintEngine *pe); |
| 369 | void setup(const QBrush &brush, int alpha, QPainter::CompositionMode compositionMode); |
| 370 | void setupMatrix(const QTransform &matrix, int bilinear); |
| 371 | void initTexture(const QImage *image, int alpha, QTextureData::Type = QTextureData::Plain, const QRect &sourceRect = QRect()); |
| 372 | void adjustSpanMethods(); |
| 373 | }; |
| 374 | |
| 375 | static inline uint qt_gradient_clamp(const QGradientData *data, int ipos) |
| 376 | { |
| 377 | if (ipos < 0 || ipos >= GRADIENT_STOPTABLE_SIZE) { |
| 378 | if (data->spread == QGradient::RepeatSpread) { |
| 379 | ipos = ipos % GRADIENT_STOPTABLE_SIZE; |
| 380 | ipos = ipos < 0 ? GRADIENT_STOPTABLE_SIZE + ipos : ipos; |
| 381 | } else if (data->spread == QGradient::ReflectSpread) { |
| 382 | const int limit = GRADIENT_STOPTABLE_SIZE * 2; |
| 383 | ipos = ipos % limit; |
| 384 | ipos = ipos < 0 ? limit + ipos : ipos; |
| 385 | ipos = ipos >= GRADIENT_STOPTABLE_SIZE ? limit - 1 - ipos : ipos; |
| 386 | } else { |
| 387 | if (ipos < 0) |
| 388 | ipos = 0; |
| 389 | else if (ipos >= GRADIENT_STOPTABLE_SIZE) |
| 390 | ipos = GRADIENT_STOPTABLE_SIZE-1; |
| 391 | } |
| 392 | } |
| 393 | |
| 394 | Q_ASSERT(ipos >= 0); |
| 395 | Q_ASSERT(ipos < GRADIENT_STOPTABLE_SIZE); |
| 396 | |
| 397 | return ipos; |
| 398 | } |
| 399 | |
| 400 | static inline uint qt_gradient_pixel(const QGradientData *data, qreal pos) |
| 401 | { |
| 402 | int ipos = int(pos * (GRADIENT_STOPTABLE_SIZE - 1) + qreal(0.5)); |
| 403 | return data->colorTable32[qt_gradient_clamp(data, ipos)]; |
| 404 | } |
| 405 | |
| 406 | #if QT_CONFIG(raster_64bit) |
| 407 | static inline const QRgba64& qt_gradient_pixel64(const QGradientData *data, qreal pos) |
| 408 | { |
| 409 | int ipos = int(pos * (GRADIENT_STOPTABLE_SIZE - 1) + qreal(0.5)); |
| 410 | return data->colorTable64[qt_gradient_clamp(data, ipos)]; |
| 411 | } |
| 412 | #endif |
| 413 | |
| 414 | static inline qreal qRadialDeterminant(qreal a, qreal b, qreal c) |
| 415 | { |
| 416 | return (b * b) - (4 * a * c); |
| 417 | } |
| 418 | |
| 419 | template <class RadialFetchFunc, typename BlendType> static |
| 420 | const BlendType * QT_FASTCALL qt_fetch_radial_gradient_template(BlendType *buffer, const Operator *op, |
| 421 | const QSpanData *data, int y, int x, int length) |
| 422 | { |
| 423 | // avoid division by zero |
| 424 | if (qFuzzyIsNull(op->radial.a)) { |
| 425 | RadialFetchFunc::memfill(buffer, RadialFetchFunc::null(), length); |
| 426 | return buffer; |
| 427 | } |
| 428 | |
| 429 | const BlendType *b = buffer; |
| 430 | qreal rx = data->m21 * (y + qreal(0.5)) |
| 431 | + data->dx + data->m11 * (x + qreal(0.5)); |
| 432 | qreal ry = data->m22 * (y + qreal(0.5)) |
| 433 | + data->dy + data->m12 * (x + qreal(0.5)); |
| 434 | bool affine = !data->m13 && !data->m23; |
| 435 | |
| 436 | BlendType *end = buffer + length; |
| 437 | if (affine) { |
| 438 | rx -= data->gradient.radial.focal.x; |
| 439 | ry -= data->gradient.radial.focal.y; |
| 440 | |
| 441 | qreal inv_a = 1 / qreal(2 * op->radial.a); |
| 442 | |
| 443 | const qreal delta_rx = data->m11; |
| 444 | const qreal delta_ry = data->m12; |
| 445 | |
| 446 | qreal b = 2*(op->radial.dr*data->gradient.radial.focal.radius + rx * op->radial.dx + ry * op->radial.dy); |
| 447 | qreal delta_b = 2*(delta_rx * op->radial.dx + delta_ry * op->radial.dy); |
| 448 | const qreal b_delta_b = 2 * b * delta_b; |
| 449 | const qreal delta_b_delta_b = 2 * delta_b * delta_b; |
| 450 | |
| 451 | const qreal bb = b * b; |
| 452 | const qreal delta_bb = delta_b * delta_b; |
| 453 | |
| 454 | b *= inv_a; |
| 455 | delta_b *= inv_a; |
| 456 | |
| 457 | const qreal rxrxryry = rx * rx + ry * ry; |
| 458 | const qreal delta_rxrxryry = delta_rx * delta_rx + delta_ry * delta_ry; |
| 459 | const qreal rx_plus_ry = 2*(rx * delta_rx + ry * delta_ry); |
| 460 | const qreal delta_rx_plus_ry = 2 * delta_rxrxryry; |
| 461 | |
| 462 | inv_a *= inv_a; |
| 463 | |
| 464 | qreal det = (bb - 4 * op->radial.a * (op->radial.sqrfr - rxrxryry)) * inv_a; |
| 465 | qreal delta_det = (b_delta_b + delta_bb + 4 * op->radial.a * (rx_plus_ry + delta_rxrxryry)) * inv_a; |
| 466 | const qreal delta_delta_det = (delta_b_delta_b + 4 * op->radial.a * delta_rx_plus_ry) * inv_a; |
| 467 | |
| 468 | RadialFetchFunc::fetch(buffer, end, op, data, det, delta_det, delta_delta_det, b, delta_b); |
| 469 | } else { |
| 470 | qreal rw = data->m23 * (y + qreal(0.5)) |
| 471 | + data->m33 + data->m13 * (x + qreal(0.5)); |
| 472 | |
| 473 | while (buffer < end) { |
| 474 | if (rw == 0) { |
| 475 | *buffer = 0; |
| 476 | } else { |
| 477 | qreal invRw = 1 / rw; |
| 478 | qreal gx = rx * invRw - data->gradient.radial.focal.x; |
| 479 | qreal gy = ry * invRw - data->gradient.radial.focal.y; |
| 480 | qreal b = 2*(op->radial.dr*data->gradient.radial.focal.radius + gx*op->radial.dx + gy*op->radial.dy); |
| 481 | qreal det = qRadialDeterminant(op->radial.a, b, op->radial.sqrfr - (gx*gx + gy*gy)); |
| 482 | |
| 483 | BlendType result = RadialFetchFunc::null(); |
| 484 | if (det >= 0) { |
| 485 | qreal detSqrt = qSqrt(det); |
| 486 | |
| 487 | qreal s0 = (-b - detSqrt) * op->radial.inv2a; |
| 488 | qreal s1 = (-b + detSqrt) * op->radial.inv2a; |
| 489 | |
| 490 | qreal s = qMax(s0, s1); |
| 491 | |
| 492 | if (data->gradient.radial.focal.radius + op->radial.dr * s >= 0) |
| 493 | result = RadialFetchFunc::fetchSingle(data->gradient, s); |
| 494 | } |
| 495 | |
| 496 | *buffer = result; |
| 497 | } |
| 498 | |
| 499 | rx += data->m11; |
| 500 | ry += data->m12; |
| 501 | rw += data->m13; |
| 502 | |
| 503 | ++buffer; |
| 504 | } |
| 505 | } |
| 506 | |
| 507 | return b; |
| 508 | } |
| 509 | |
| 510 | template <class Simd> |
| 511 | class QRadialFetchSimd |
| 512 | { |
| 513 | public: |
| 514 | static uint null() { return 0; } |
| 515 | static uint fetchSingle(const QGradientData& gradient, qreal v) |
| 516 | { |
| 517 | return qt_gradient_pixel(&gradient, v); |
| 518 | } |
| 519 | static void memfill(uint *buffer, uint fill, int length) |
| 520 | { |
| 521 | qt_memfill32(buffer, fill, length); |
| 522 | } |
| 523 | static void fetch(uint *buffer, uint *end, const Operator *op, const QSpanData *data, qreal det, |
| 524 | qreal delta_det, qreal delta_delta_det, qreal b, qreal delta_b) |
| 525 | { |
| 526 | typename Simd::Vect_buffer_f det_vec; |
| 527 | typename Simd::Vect_buffer_f delta_det4_vec; |
| 528 | typename Simd::Vect_buffer_f b_vec; |
| 529 | |
| 530 | for (int i = 0; i < 4; ++i) { |
| 531 | det_vec.f[i] = det; |
| 532 | delta_det4_vec.f[i] = 4 * delta_det; |
| 533 | b_vec.f[i] = b; |
| 534 | |
| 535 | det += delta_det; |
| 536 | delta_det += delta_delta_det; |
| 537 | b += delta_b; |
| 538 | } |
| 539 | |
| 540 | const typename Simd::Float32x4 v_delta_delta_det16 = Simd::v_dup(16 * delta_delta_det); |
| 541 | const typename Simd::Float32x4 v_delta_delta_det6 = Simd::v_dup(6 * delta_delta_det); |
| 542 | const typename Simd::Float32x4 v_delta_b4 = Simd::v_dup(4 * delta_b); |
| 543 | |
| 544 | const typename Simd::Float32x4 v_r0 = Simd::v_dup(data->gradient.radial.focal.radius); |
| 545 | const typename Simd::Float32x4 v_dr = Simd::v_dup(op->radial.dr); |
| 546 | |
| 547 | #if defined(__ARM_NEON__) |
| 548 | // NEON doesn't have SIMD sqrt, but uses rsqrt instead that can't be taken of 0. |
| 549 | const typename Simd::Float32x4 v_min = Simd::v_dup(std::numeric_limits<float>::epsilon()); |
| 550 | #else |
| 551 | const typename Simd::Float32x4 v_min = Simd::v_dup(0.0f); |
| 552 | #endif |
| 553 | const typename Simd::Float32x4 v_max = Simd::v_dup(float(GRADIENT_STOPTABLE_SIZE-1)); |
| 554 | const typename Simd::Float32x4 v_half = Simd::v_dup(0.5f); |
| 555 | |
| 556 | const typename Simd::Int32x4 v_repeat_mask = Simd::v_dup(~(uint(0xffffff) << GRADIENT_STOPTABLE_SIZE_SHIFT)); |
| 557 | const typename Simd::Int32x4 v_reflect_mask = Simd::v_dup(~(uint(0xffffff) << (GRADIENT_STOPTABLE_SIZE_SHIFT+1))); |
| 558 | |
| 559 | const typename Simd::Int32x4 v_reflect_limit = Simd::v_dup(2 * GRADIENT_STOPTABLE_SIZE - 1); |
| 560 | |
| 561 | const int extended_mask = op->radial.extended ? 0x0 : ~0x0; |
| 562 | |
| 563 | #define FETCH_RADIAL_LOOP_PROLOGUE \ |
| 564 | while (buffer < end) { \ |
| 565 | typename Simd::Vect_buffer_i v_buffer_mask; \ |
| 566 | v_buffer_mask.v = Simd::v_greaterOrEqual(det_vec.v, v_min); \ |
| 567 | const typename Simd::Float32x4 v_index_local = Simd::v_sub(Simd::v_sqrt(Simd::v_max(v_min, det_vec.v)), b_vec.v); \ |
| 568 | const typename Simd::Float32x4 v_index = Simd::v_add(Simd::v_mul(v_index_local, v_max), v_half); \ |
| 569 | 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)); \ |
| 570 | typename Simd::Vect_buffer_i index_vec; |
| 571 | #define FETCH_RADIAL_LOOP_CLAMP_REPEAT \ |
| 572 | index_vec.v = Simd::v_and(v_repeat_mask, Simd::v_toInt(v_index)); |
| 573 | #define FETCH_RADIAL_LOOP_CLAMP_REFLECT \ |
| 574 | const typename Simd::Int32x4 v_index_i = Simd::v_and(v_reflect_mask, Simd::v_toInt(v_index)); \ |
| 575 | const typename Simd::Int32x4 v_index_i_inv = Simd::v_sub(v_reflect_limit, v_index_i); \ |
| 576 | index_vec.v = Simd::v_min_16(v_index_i, v_index_i_inv); |
| 577 | #define FETCH_RADIAL_LOOP_CLAMP_PAD \ |
| 578 | index_vec.v = Simd::v_toInt(Simd::v_min(v_max, Simd::v_max(v_min, v_index))); |
| 579 | #define FETCH_RADIAL_LOOP_EPILOGUE \ |
| 580 | det_vec.v = Simd::v_add(Simd::v_add(det_vec.v, delta_det4_vec.v), v_delta_delta_det6); \ |
| 581 | delta_det4_vec.v = Simd::v_add(delta_det4_vec.v, v_delta_delta_det16); \ |
| 582 | b_vec.v = Simd::v_add(b_vec.v, v_delta_b4); \ |
| 583 | for (int i = 0; i < 4; ++i) \ |
| 584 | *buffer++ = (extended_mask | v_buffer_mask.i[i]) & data->gradient.colorTable32[index_vec.i[i]]; \ |
| 585 | } |
| 586 | |
| 587 | #define FETCH_RADIAL_LOOP(FETCH_RADIAL_LOOP_CLAMP) \ |
| 588 | FETCH_RADIAL_LOOP_PROLOGUE \ |
| 589 | FETCH_RADIAL_LOOP_CLAMP \ |
| 590 | FETCH_RADIAL_LOOP_EPILOGUE |
| 591 | |
| 592 | switch (data->gradient.spread) { |
| 593 | case QGradient::RepeatSpread: |
| 594 | FETCH_RADIAL_LOOP(FETCH_RADIAL_LOOP_CLAMP_REPEAT) |
| 595 | break; |
| 596 | case QGradient::ReflectSpread: |
| 597 | FETCH_RADIAL_LOOP(FETCH_RADIAL_LOOP_CLAMP_REFLECT) |
| 598 | break; |
| 599 | case QGradient::PadSpread: |
| 600 | FETCH_RADIAL_LOOP(FETCH_RADIAL_LOOP_CLAMP_PAD) |
| 601 | break; |
| 602 | default: |
| 603 | Q_UNREACHABLE(); |
| 604 | } |
| 605 | } |
| 606 | }; |
| 607 | |
| 608 | static inline uint INTERPOLATE_PIXEL_255(uint x, uint a, uint y, uint b) { |
| 609 | uint t = (x & 0xff00ff) * a + (y & 0xff00ff) * b; |
| 610 | t = (t + ((t >> 8) & 0xff00ff) + 0x800080) >> 8; |
| 611 | t &= 0xff00ff; |
| 612 | |
| 613 | x = ((x >> 8) & 0xff00ff) * a + ((y >> 8) & 0xff00ff) * b; |
| 614 | x = (x + ((x >> 8) & 0xff00ff) + 0x800080); |
| 615 | x &= 0xff00ff00; |
| 616 | x |= t; |
| 617 | return x; |
| 618 | } |
| 619 | |
| 620 | #if Q_PROCESSOR_WORDSIZE == 8 // 64-bit versions |
| 621 | |
| 622 | static inline uint INTERPOLATE_PIXEL_256(uint x, uint a, uint y, uint b) { |
| 623 | quint64 t = (((quint64(x)) | ((quint64(x)) << 24)) & 0x00ff00ff00ff00ff) * a; |
| 624 | t += (((quint64(y)) | ((quint64(y)) << 24)) & 0x00ff00ff00ff00ff) * b; |
| 625 | t >>= 8; |
| 626 | t &= 0x00ff00ff00ff00ff; |
| 627 | return (uint(t)) | (uint(t >> 24)); |
| 628 | } |
| 629 | |
| 630 | static inline uint BYTE_MUL(uint x, uint a) { |
| 631 | quint64 t = (((quint64(x)) | ((quint64(x)) << 24)) & 0x00ff00ff00ff00ff) * a; |
| 632 | t = (t + ((t >> 8) & 0xff00ff00ff00ff) + 0x80008000800080) >> 8; |
| 633 | t &= 0x00ff00ff00ff00ff; |
| 634 | return (uint(t)) | (uint(t >> 24)); |
| 635 | } |
| 636 | |
| 637 | #else // 32-bit versions |
| 638 | |
| 639 | static inline uint INTERPOLATE_PIXEL_256(uint x, uint a, uint y, uint b) { |
| 640 | uint t = (x & 0xff00ff) * a + (y & 0xff00ff) * b; |
| 641 | t >>= 8; |
| 642 | t &= 0xff00ff; |
| 643 | |
| 644 | x = ((x >> 8) & 0xff00ff) * a + ((y >> 8) & 0xff00ff) * b; |
| 645 | x &= 0xff00ff00; |
| 646 | x |= t; |
| 647 | return x; |
| 648 | } |
| 649 | |
| 650 | static inline uint BYTE_MUL(uint x, uint a) { |
| 651 | uint t = (x & 0xff00ff) * a; |
| 652 | t = (t + ((t >> 8) & 0xff00ff) + 0x800080) >> 8; |
| 653 | t &= 0xff00ff; |
| 654 | |
| 655 | x = ((x >> 8) & 0xff00ff) * a; |
| 656 | x = (x + ((x >> 8) & 0xff00ff) + 0x800080); |
| 657 | x &= 0xff00ff00; |
| 658 | x |= t; |
| 659 | return x; |
| 660 | } |
| 661 | #endif |
| 662 | |
| 663 | static inline void blend_pixel(quint32 &dst, const quint32 src) |
| 664 | { |
| 665 | if (src >= 0xff000000) |
| 666 | dst = src; |
| 667 | else if (src != 0) |
| 668 | dst = src + BYTE_MUL(dst, qAlpha(~src)); |
| 669 | } |
| 670 | |
| 671 | static inline void blend_pixel(quint32 &dst, const quint32 src, const int const_alpha) |
| 672 | { |
| 673 | if (const_alpha == 255) |
| 674 | return blend_pixel(dst, src); |
| 675 | if (src != 0) { |
| 676 | const quint32 s = BYTE_MUL(src, const_alpha); |
| 677 | dst = s + BYTE_MUL(dst, qAlpha(~s)); |
| 678 | } |
| 679 | } |
| 680 | |
| 681 | #if defined(__SSE2__) |
| 682 | static inline uint Q_DECL_VECTORCALL interpolate_4_pixels_sse2(__m128i vt, __m128i vb, uint distx, uint disty) |
| 683 | { |
| 684 | // First interpolate top and bottom pixels in parallel. |
| 685 | vt = _mm_unpacklo_epi8(vt, _mm_setzero_si128()); |
| 686 | vb = _mm_unpacklo_epi8(vb, _mm_setzero_si128()); |
| 687 | vt = _mm_mullo_epi16(vt, _mm_set1_epi16(256 - disty)); |
| 688 | vb = _mm_mullo_epi16(vb, _mm_set1_epi16(disty)); |
| 689 | __m128i vlr = _mm_add_epi16(vt, vb); |
| 690 | vlr = _mm_srli_epi16(vlr, 8); |
| 691 | // vlr now contains the result of the first two interpolate calls vlr = unpacked((xright << 64) | xleft) |
| 692 | |
| 693 | // Now the last interpolate between left and right.. |
| 694 | const __m128i vidistx = _mm_shufflelo_epi16(_mm_cvtsi32_si128(256 - distx), _MM_SHUFFLE(0, 0, 0, 0)); |
| 695 | const __m128i vdistx = _mm_shufflelo_epi16(_mm_cvtsi32_si128(distx), _MM_SHUFFLE(0, 0, 0, 0)); |
| 696 | const __m128i vmulx = _mm_unpacklo_epi16(vidistx, vdistx); |
| 697 | vlr = _mm_unpacklo_epi16(vlr, _mm_srli_si128(vlr, 8)); |
| 698 | // vlr now contains the colors of left and right interleaved { la, ra, lr, rr, lg, rg, lb, rb } |
| 699 | vlr = _mm_madd_epi16(vlr, vmulx); // Multiply and horizontal add. |
| 700 | vlr = _mm_srli_epi32(vlr, 8); |
| 701 | vlr = _mm_packs_epi32(vlr, vlr); |
| 702 | vlr = _mm_packus_epi16(vlr, vlr); |
| 703 | return _mm_cvtsi128_si32(vlr); |
| 704 | } |
| 705 | |
| 706 | static inline uint interpolate_4_pixels(uint tl, uint tr, uint bl, uint br, uint distx, uint disty) |
| 707 | { |
| 708 | __m128i vt = _mm_unpacklo_epi32(_mm_cvtsi32_si128(tl), _mm_cvtsi32_si128(tr)); |
| 709 | __m128i vb = _mm_unpacklo_epi32(_mm_cvtsi32_si128(bl), _mm_cvtsi32_si128(br)); |
| 710 | return interpolate_4_pixels_sse2(vt, vb, distx, disty); |
| 711 | } |
| 712 | |
| 713 | static inline uint interpolate_4_pixels(const uint t[], const uint b[], uint distx, uint disty) |
| 714 | { |
| 715 | __m128i vt = _mm_loadl_epi64((const __m128i*)t); |
| 716 | __m128i vb = _mm_loadl_epi64((const __m128i*)b); |
| 717 | return interpolate_4_pixels_sse2(vt, vb, distx, disty); |
| 718 | } |
| 719 | |
| 720 | static constexpr inline bool hasFastInterpolate4() { return true; } |
| 721 | |
| 722 | #elif defined(__ARM_NEON__) |
| 723 | static inline uint interpolate_4_pixels_neon(uint32x2_t vt32, uint32x2_t vb32, uint distx, uint disty) |
| 724 | { |
| 725 | uint16x8_t vt16 = vmovl_u8(vreinterpret_u8_u32(vt32)); |
| 726 | uint16x8_t vb16 = vmovl_u8(vreinterpret_u8_u32(vb32)); |
| 727 | vt16 = vmulq_n_u16(vt16, 256 - disty); |
| 728 | vt16 = vmlaq_n_u16(vt16, vb16, disty); |
| 729 | vt16 = vshrq_n_u16(vt16, 8); |
| 730 | uint16x4_t vl16 = vget_low_u16(vt16); |
| 731 | uint16x4_t vr16 = vget_high_u16(vt16); |
| 732 | vl16 = vmul_n_u16(vl16, 256 - distx); |
| 733 | vl16 = vmla_n_u16(vl16, vr16, distx); |
| 734 | vl16 = vshr_n_u16(vl16, 8); |
| 735 | uint8x8_t vr = vmovn_u16(vcombine_u16(vl16, vl16)); |
| 736 | return vget_lane_u32(vreinterpret_u32_u8(vr), 0); |
| 737 | } |
| 738 | |
| 739 | static inline uint interpolate_4_pixels(uint tl, uint tr, uint bl, uint br, uint distx, uint disty) |
| 740 | { |
| 741 | uint32x2_t vt32 = vmov_n_u32(tl); |
| 742 | uint32x2_t vb32 = vmov_n_u32(bl); |
| 743 | vt32 = vset_lane_u32(tr, vt32, 1); |
| 744 | vb32 = vset_lane_u32(br, vb32, 1); |
| 745 | return interpolate_4_pixels_neon(vt32, vb32, distx, disty); |
| 746 | } |
| 747 | |
| 748 | static inline uint interpolate_4_pixels(const uint t[], const uint b[], uint distx, uint disty) |
| 749 | { |
| 750 | uint32x2_t vt32 = vld1_u32(t); |
| 751 | uint32x2_t vb32 = vld1_u32(b); |
| 752 | return interpolate_4_pixels_neon(vt32, vb32, distx, disty); |
| 753 | } |
| 754 | |
| 755 | static constexpr inline bool hasFastInterpolate4() { return true; } |
| 756 | |
| 757 | #else |
| 758 | static inline uint interpolate_4_pixels(uint tl, uint tr, uint bl, uint br, uint distx, uint disty) |
| 759 | { |
| 760 | uint idistx = 256 - distx; |
| 761 | uint idisty = 256 - disty; |
| 762 | uint xtop = INTERPOLATE_PIXEL_256(tl, idistx, tr, distx); |
| 763 | uint xbot = INTERPOLATE_PIXEL_256(bl, idistx, br, distx); |
| 764 | return INTERPOLATE_PIXEL_256(xtop, idisty, xbot, disty); |
| 765 | } |
| 766 | |
| 767 | static inline uint interpolate_4_pixels(const uint t[], const uint b[], uint distx, uint disty) |
| 768 | { |
| 769 | return interpolate_4_pixels(t[0], t[1], b[0], b[1], distx, disty); |
| 770 | } |
| 771 | |
| 772 | static constexpr inline bool hasFastInterpolate4() { return false; } |
| 773 | |
| 774 | #endif |
| 775 | |
| 776 | static inline QRgba64 multiplyAlpha256(QRgba64 rgba64, uint alpha256) |
| 777 | { |
| 778 | return QRgba64::fromRgba64((rgba64.red() * alpha256) >> 8, |
| 779 | (rgba64.green() * alpha256) >> 8, |
| 780 | (rgba64.blue() * alpha256) >> 8, |
| 781 | (rgba64.alpha() * alpha256) >> 8); |
| 782 | } |
| 783 | static inline QRgba64 interpolate256(QRgba64 x, uint alpha1, QRgba64 y, uint alpha2) |
| 784 | { |
| 785 | return QRgba64::fromRgba64(multiplyAlpha256(x, alpha1) + multiplyAlpha256(y, alpha2)); |
| 786 | } |
| 787 | |
| 788 | #ifdef __SSE2__ |
| 789 | static inline QRgba64 interpolate_4_pixels_rgb64(const QRgba64 t[], const QRgba64 b[], uint distx, uint disty) |
| 790 | { |
| 791 | __m128i vt = _mm_loadu_si128((const __m128i*)t); |
| 792 | if (disty) { |
| 793 | __m128i vb = _mm_loadu_si128((const __m128i*)b); |
| 794 | vt = _mm_mulhi_epu16(vt, _mm_set1_epi16(0x10000 - disty)); |
| 795 | vb = _mm_mulhi_epu16(vb, _mm_set1_epi16(disty)); |
| 796 | vt = _mm_add_epi16(vt, vb); |
| 797 | } |
| 798 | if (distx) { |
| 799 | const __m128i vdistx = _mm_shufflelo_epi16(_mm_cvtsi32_si128(distx), _MM_SHUFFLE(0, 0, 0, 0)); |
| 800 | const __m128i vidistx = _mm_shufflelo_epi16(_mm_cvtsi32_si128(0x10000 - distx), _MM_SHUFFLE(0, 0, 0, 0)); |
| 801 | vt = _mm_mulhi_epu16(vt, _mm_unpacklo_epi64(vidistx, vdistx)); |
| 802 | vt = _mm_add_epi16(vt, _mm_srli_si128(vt, 8)); |
| 803 | } |
| 804 | #ifdef Q_PROCESSOR_X86_64 |
| 805 | return QRgba64::fromRgba64(_mm_cvtsi128_si64(vt)); |
| 806 | #else |
| 807 | QRgba64 out; |
| 808 | _mm_storel_epi64((__m128i*)&out, vt); |
| 809 | return out; |
| 810 | #endif // Q_PROCESSOR_X86_64 |
| 811 | } |
| 812 | #elif defined(__ARM_NEON__) |
| 813 | static inline QRgba64 interpolate_4_pixels_rgb64(const QRgba64 t[], const QRgba64 b[], uint distx, uint disty) |
| 814 | { |
| 815 | uint64x1x2_t vt = vld2_u64(reinterpret_cast<const uint64_t *>(t)); |
| 816 | if (disty) { |
| 817 | uint64x1x2_t vb = vld2_u64(reinterpret_cast<const uint64_t *>(b)); |
| 818 | uint32x4_t vt0 = vmull_n_u16(vreinterpret_u16_u64(vt.val[0]), 0x10000 - disty); |
| 819 | uint32x4_t vt1 = vmull_n_u16(vreinterpret_u16_u64(vt.val[1]), 0x10000 - disty); |
| 820 | vt0 = vmlal_n_u16(vt0, vreinterpret_u16_u64(vb.val[0]), disty); |
| 821 | vt1 = vmlal_n_u16(vt1, vreinterpret_u16_u64(vb.val[1]), disty); |
| 822 | vt.val[0] = vreinterpret_u64_u16(vshrn_n_u32(vt0, 16)); |
| 823 | vt.val[1] = vreinterpret_u64_u16(vshrn_n_u32(vt1, 16)); |
| 824 | } |
| 825 | if (distx) { |
| 826 | uint32x4_t vt0 = vmull_n_u16(vreinterpret_u16_u64(vt.val[0]), 0x10000 - distx); |
| 827 | vt0 = vmlal_n_u16(vt0, vreinterpret_u16_u64(vt.val[1]), distx); |
| 828 | vt.val[0] = vreinterpret_u64_u16(vshrn_n_u32(vt0, 16)); |
| 829 | } |
| 830 | QRgba64 out; |
| 831 | vst1_u64(reinterpret_cast<uint64_t *>(&out), vt.val[0]); |
| 832 | return out; |
| 833 | } |
| 834 | #else |
| 835 | static inline QRgba64 interpolate_4_pixels_rgb64(const QRgba64 t[], const QRgba64 b[], uint distx, uint disty) |
| 836 | { |
| 837 | const uint dx = distx>>8; |
| 838 | const uint dy = disty>>8; |
| 839 | const uint idx = 256 - dx; |
| 840 | const uint idy = 256 - dy; |
| 841 | QRgba64 xtop = interpolate256(t[0], idx, t[1], dx); |
| 842 | QRgba64 xbot = interpolate256(b[0], idx, b[1], dx); |
| 843 | return interpolate256(xtop, idy, xbot, dy); |
| 844 | } |
| 845 | #endif // __SSE2__ |
| 846 | |
| 847 | static inline uint BYTE_MUL_RGB16(uint x, uint a) { |
| 848 | a += 1; |
| 849 | uint t = (((x & 0x07e0)*a) >> 8) & 0x07e0; |
| 850 | t |= (((x & 0xf81f)*(a>>2)) >> 6) & 0xf81f; |
| 851 | return t; |
| 852 | } |
| 853 | |
| 854 | static inline uint BYTE_MUL_RGB16_32(uint x, uint a) { |
| 855 | uint t = (((x & 0xf81f07e0) >> 5)*a) & 0xf81f07e0; |
| 856 | t |= (((x & 0x07e0f81f)*a) >> 5) & 0x07e0f81f; |
| 857 | return t; |
| 858 | } |
| 859 | |
| 860 | // qt_div_255 is a fast rounded division by 255 using an approximation that is accurate for all positive 16-bit integers |
| 861 | static constexpr inline int qt_div_255(int x) { return (x + (x>>8) + 0x80) >> 8; } |
| 862 | static constexpr inline uint qt_div_257_floor(uint x) { return (x - (x >> 8)) >> 8; } |
| 863 | static constexpr inline uint qt_div_257(uint x) { return qt_div_257_floor(x + 128); } |
| 864 | static constexpr inline uint qt_div_65535(uint x) { return (x + (x>>16) + 0x8000U) >> 16; } |
| 865 | |
| 866 | template <class T> inline void qt_memfill_template(T *dest, T color, qsizetype count) |
| 867 | { |
| 868 | if (!count) |
| 869 | return; |
| 870 | |
| 871 | qsizetype n = (count + 7) / 8; |
| 872 | switch (count & 0x07) |
| 873 | { |
| 874 | case 0: do { *dest++ = color; Q_FALLTHROUGH(); |
| 875 | case 7: *dest++ = color; Q_FALLTHROUGH(); |
| 876 | case 6: *dest++ = color; Q_FALLTHROUGH(); |
| 877 | case 5: *dest++ = color; Q_FALLTHROUGH(); |
| 878 | case 4: *dest++ = color; Q_FALLTHROUGH(); |
| 879 | case 3: *dest++ = color; Q_FALLTHROUGH(); |
| 880 | case 2: *dest++ = color; Q_FALLTHROUGH(); |
| 881 | case 1: *dest++ = color; |
| 882 | } while (--n > 0); |
| 883 | } |
| 884 | } |
| 885 | |
| 886 | template <class T> inline void qt_memfill(T *dest, T value, qsizetype count) |
| 887 | { |
| 888 | qt_memfill_template(dest, value, count); |
| 889 | } |
| 890 | |
| 891 | template<> inline void qt_memfill(quint64 *dest, quint64 color, qsizetype count) |
| 892 | { |
| 893 | qt_memfill64(dest, color, count); |
| 894 | } |
| 895 | |
| 896 | template<> inline void qt_memfill(quint32 *dest, quint32 color, qsizetype count) |
| 897 | { |
| 898 | qt_memfill32(dest, color, count); |
| 899 | } |
| 900 | |
| 901 | template<> inline void qt_memfill(quint24 *dest, quint24 color, qsizetype count) |
| 902 | { |
| 903 | qt_memfill24(dest, color, count); |
| 904 | } |
| 905 | |
| 906 | template<> inline void qt_memfill(quint16 *dest, quint16 color, qsizetype count) |
| 907 | { |
| 908 | qt_memfill16(dest, color, count); |
| 909 | } |
| 910 | |
| 911 | template<> inline void qt_memfill(quint8 *dest, quint8 color, qsizetype count) |
| 912 | { |
| 913 | memset(dest, color, count); |
| 914 | } |
| 915 | |
| 916 | template <class T> static |
| 917 | inline void qt_rectfill(T *dest, T value, |
| 918 | int x, int y, int width, int height, qsizetype stride) |
| 919 | { |
| 920 | char *d = reinterpret_cast<char*>(dest + x) + y * stride; |
| 921 | if (uint(stride) == (width * sizeof(T))) { |
| 922 | qt_memfill(reinterpret_cast<T*>(d), value, qsizetype(width) * height); |
| 923 | } else { |
| 924 | for (int j = 0; j < height; ++j) { |
| 925 | dest = reinterpret_cast<T*>(d); |
| 926 | qt_memfill(dest, value, width); |
| 927 | d += stride; |
| 928 | } |
| 929 | } |
| 930 | } |
| 931 | |
| 932 | inline ushort qConvertRgb32To16(uint c) |
| 933 | { |
| 934 | return (((c) >> 3) & 0x001f) |
| 935 | | (((c) >> 5) & 0x07e0) |
| 936 | | (((c) >> 8) & 0xf800); |
| 937 | } |
| 938 | |
| 939 | inline QRgb qConvertRgb16To32(uint c) |
| 940 | { |
| 941 | return 0xff000000 |
| 942 | | ((((c) << 3) & 0xf8) | (((c) >> 2) & 0x7)) |
| 943 | | ((((c) << 5) & 0xfc00) | (((c) >> 1) & 0x300)) |
| 944 | | ((((c) << 8) & 0xf80000) | (((c) << 3) & 0x70000)); |
| 945 | } |
| 946 | |
| 947 | const uint qt_bayer_matrix[16][16] = { |
| 948 | { 0x1, 0xc0, 0x30, 0xf0, 0xc, 0xcc, 0x3c, 0xfc, |
| 949 | 0x3, 0xc3, 0x33, 0xf3, 0xf, 0xcf, 0x3f, 0xff}, |
| 950 | { 0x80, 0x40, 0xb0, 0x70, 0x8c, 0x4c, 0xbc, 0x7c, |
| 951 | 0x83, 0x43, 0xb3, 0x73, 0x8f, 0x4f, 0xbf, 0x7f}, |
| 952 | { 0x20, 0xe0, 0x10, 0xd0, 0x2c, 0xec, 0x1c, 0xdc, |
| 953 | 0x23, 0xe3, 0x13, 0xd3, 0x2f, 0xef, 0x1f, 0xdf}, |
| 954 | { 0xa0, 0x60, 0x90, 0x50, 0xac, 0x6c, 0x9c, 0x5c, |
| 955 | 0xa3, 0x63, 0x93, 0x53, 0xaf, 0x6f, 0x9f, 0x5f}, |
| 956 | { 0x8, 0xc8, 0x38, 0xf8, 0x4, 0xc4, 0x34, 0xf4, |
| 957 | 0xb, 0xcb, 0x3b, 0xfb, 0x7, 0xc7, 0x37, 0xf7}, |
| 958 | { 0x88, 0x48, 0xb8, 0x78, 0x84, 0x44, 0xb4, 0x74, |
| 959 | 0x8b, 0x4b, 0xbb, 0x7b, 0x87, 0x47, 0xb7, 0x77}, |
| 960 | { 0x28, 0xe8, 0x18, 0xd8, 0x24, 0xe4, 0x14, 0xd4, |
| 961 | 0x2b, 0xeb, 0x1b, 0xdb, 0x27, 0xe7, 0x17, 0xd7}, |
| 962 | { 0xa8, 0x68, 0x98, 0x58, 0xa4, 0x64, 0x94, 0x54, |
| 963 | 0xab, 0x6b, 0x9b, 0x5b, 0xa7, 0x67, 0x97, 0x57}, |
| 964 | { 0x2, 0xc2, 0x32, 0xf2, 0xe, 0xce, 0x3e, 0xfe, |
| 965 | 0x1, 0xc1, 0x31, 0xf1, 0xd, 0xcd, 0x3d, 0xfd}, |
| 966 | { 0x82, 0x42, 0xb2, 0x72, 0x8e, 0x4e, 0xbe, 0x7e, |
| 967 | 0x81, 0x41, 0xb1, 0x71, 0x8d, 0x4d, 0xbd, 0x7d}, |
| 968 | { 0x22, 0xe2, 0x12, 0xd2, 0x2e, 0xee, 0x1e, 0xde, |
| 969 | 0x21, 0xe1, 0x11, 0xd1, 0x2d, 0xed, 0x1d, 0xdd}, |
| 970 | { 0xa2, 0x62, 0x92, 0x52, 0xae, 0x6e, 0x9e, 0x5e, |
| 971 | 0xa1, 0x61, 0x91, 0x51, 0xad, 0x6d, 0x9d, 0x5d}, |
| 972 | { 0xa, 0xca, 0x3a, 0xfa, 0x6, 0xc6, 0x36, 0xf6, |
| 973 | 0x9, 0xc9, 0x39, 0xf9, 0x5, 0xc5, 0x35, 0xf5}, |
| 974 | { 0x8a, 0x4a, 0xba, 0x7a, 0x86, 0x46, 0xb6, 0x76, |
| 975 | 0x89, 0x49, 0xb9, 0x79, 0x85, 0x45, 0xb5, 0x75}, |
| 976 | { 0x2a, 0xea, 0x1a, 0xda, 0x26, 0xe6, 0x16, 0xd6, |
| 977 | 0x29, 0xe9, 0x19, 0xd9, 0x25, 0xe5, 0x15, 0xd5}, |
| 978 | { 0xaa, 0x6a, 0x9a, 0x5a, 0xa6, 0x66, 0x96, 0x56, |
| 979 | 0xa9, 0x69, 0x99, 0x59, 0xa5, 0x65, 0x95, 0x55} |
| 980 | }; |
| 981 | |
| 982 | #define ARGB_COMBINE_ALPHA(argb, alpha) \ |
| 983 | ((((argb >> 24) * alpha) >> 8) << 24) | (argb & 0x00ffffff) |
| 984 | |
| 985 | |
| 986 | #if Q_PROCESSOR_WORDSIZE == 8 // 64-bit versions |
| 987 | #define AMIX(mask) (qMin(((quint64(s)&mask) + (quint64(d)&mask)), quint64(mask))) |
| 988 | #define MIX(mask) (qMin(((quint64(s)&mask) + (quint64(d)&mask)), quint64(mask))) |
| 989 | #else // 32 bits |
| 990 | // The mask for alpha can overflow over 32 bits |
| 991 | #define AMIX(mask) quint32(qMin(((quint64(s)&mask) + (quint64(d)&mask)), quint64(mask))) |
| 992 | #define MIX(mask) (qMin(((quint32(s)&mask) + (quint32(d)&mask)), quint32(mask))) |
| 993 | #endif |
| 994 | |
| 995 | inline uint comp_func_Plus_one_pixel_const_alpha(uint d, const uint s, const uint const_alpha, const uint one_minus_const_alpha) |
| 996 | { |
| 997 | const uint result = uint(AMIX(AMASK) | MIX(RMASK) | MIX(GMASK) | MIX(BMASK)); |
| 998 | return INTERPOLATE_PIXEL_255(result, const_alpha, d, one_minus_const_alpha); |
| 999 | } |
| 1000 | |
| 1001 | inline uint comp_func_Plus_one_pixel(uint d, const uint s) |
| 1002 | { |
| 1003 | const uint result = uint(AMIX(AMASK) | MIX(RMASK) | MIX(GMASK) | MIX(BMASK)); |
| 1004 | return result; |
| 1005 | } |
| 1006 | |
| 1007 | #undef MIX |
| 1008 | #undef AMIX |
| 1009 | |
| 1010 | // must be multiple of 4 for easier SIMD implementations |
| 1011 | static constexpr int BufferSize = 2048; |
| 1012 | |
| 1013 | // A buffer of intermediate results used by simple bilinear scaling. |
| 1014 | struct IntermediateBuffer |
| 1015 | { |
| 1016 | // The idea is first to do the interpolation between the row s1 and the row s2 |
| 1017 | // into this intermediate buffer, then later interpolate between two pixel of this buffer. |
| 1018 | // |
| 1019 | // buffer_rb is a buffer of red-blue component of the pixel, in the form 0x00RR00BB |
| 1020 | // buffer_ag is the alpha-green component of the pixel, in the form 0x00AA00GG |
| 1021 | // +1 for the last pixel to interpolate with, and +1 for rounding errors. |
| 1022 | quint32 buffer_rb[BufferSize+2]; |
| 1023 | quint32 buffer_ag[BufferSize+2]; |
| 1024 | }; |
| 1025 | |
| 1026 | template <QPixelLayout::BPP bpp> |
| 1027 | inline uint QT_FASTCALL qFetchPixel(const uchar *, int) |
| 1028 | { |
| 1029 | Q_UNREACHABLE(); |
| 1030 | return 0; |
| 1031 | } |
| 1032 | |
| 1033 | template <> |
| 1034 | inline uint QT_FASTCALL qFetchPixel<QPixelLayout::BPP1LSB>(const uchar *src, int index) |
| 1035 | { |
| 1036 | return (src[index >> 3] >> (index & 7)) & 1; |
| 1037 | } |
| 1038 | |
| 1039 | template <> |
| 1040 | inline uint QT_FASTCALL qFetchPixel<QPixelLayout::BPP1MSB>(const uchar *src, int index) |
| 1041 | { |
| 1042 | return (src[index >> 3] >> (~index & 7)) & 1; |
| 1043 | } |
| 1044 | |
| 1045 | template <> |
| 1046 | inline uint QT_FASTCALL qFetchPixel<QPixelLayout::BPP8>(const uchar *src, int index) |
| 1047 | { |
| 1048 | return src[index]; |
| 1049 | } |
| 1050 | |
| 1051 | template <> |
| 1052 | inline uint QT_FASTCALL qFetchPixel<QPixelLayout::BPP16>(const uchar *src, int index) |
| 1053 | { |
| 1054 | return reinterpret_cast<const quint16 *>(src)[index]; |
| 1055 | } |
| 1056 | |
| 1057 | template <> |
| 1058 | inline uint QT_FASTCALL qFetchPixel<QPixelLayout::BPP24>(const uchar *src, int index) |
| 1059 | { |
| 1060 | return reinterpret_cast<const quint24 *>(src)[index]; |
| 1061 | } |
| 1062 | |
| 1063 | template <> |
| 1064 | inline uint QT_FASTCALL qFetchPixel<QPixelLayout::BPP32>(const uchar *src, int index) |
| 1065 | { |
| 1066 | return reinterpret_cast<const uint *>(src)[index]; |
| 1067 | } |
| 1068 | |
| 1069 | template <> |
| 1070 | inline uint QT_FASTCALL qFetchPixel<QPixelLayout::BPP64>(const uchar *src, int index) |
| 1071 | { |
| 1072 | // We have to do the conversion in fetch to fit into a 32bit uint |
| 1073 | QRgba64 c = reinterpret_cast<const QRgba64 *>(src)[index]; |
| 1074 | return c.toArgb32(); |
| 1075 | } |
| 1076 | |
| 1077 | typedef uint (QT_FASTCALL *FetchPixelFunc)(const uchar *src, int index); |
| 1078 | |
| 1079 | constexpr FetchPixelFunc qFetchPixelTable[QPixelLayout::BPPCount] = { |
| 1080 | nullptr, // BPPNone |
| 1081 | qFetchPixel<QPixelLayout::BPP1MSB>, |
| 1082 | qFetchPixel<QPixelLayout::BPP1LSB>, |
| 1083 | qFetchPixel<QPixelLayout::BPP8>, |
| 1084 | qFetchPixel<QPixelLayout::BPP16>, |
| 1085 | qFetchPixel<QPixelLayout::BPP24>, |
| 1086 | qFetchPixel<QPixelLayout::BPP32>, |
| 1087 | qFetchPixel<QPixelLayout::BPP64>, |
| 1088 | }; |
| 1089 | |
| 1090 | QT_END_NAMESPACE |
| 1091 | |
| 1092 | #endif // QDRAWHELPER_P_H |
| 1093 | |