"Fossies" - the Fresh Open Source Software Archive  

Source code changes of the file "benchmarks/kis_composition_benchmark.cpp" between
krita-5.1.3.tar.xz and krita-5.1.4.tar.xz

About: Krita is a painting program offering art tools for everyone (concept art, texture and matte painters, illustrations and comics).

kis_composition_benchmark.cpp  (krita-5.1.3.tar.xz)	:	kis_composition_benchmark.cpp  (krita-5.1.4.tar.xz)
skipping to change at line 12		skipping to change at line 12
* SPDX-FileCopyrightText: 2012 Dmitry Kazakov <dimula73@gmail.com>		* SPDX-FileCopyrightText: 2012 Dmitry Kazakov <dimula73@gmail.com>
* SPDX-FileCopyrightText: 2015 Thorsten Zachmann <zachmann@kde.org>		* SPDX-FileCopyrightText: 2015 Thorsten Zachmann <zachmann@kde.org>
* SPDX-FileCopyrightText: 2020 Mathias Wein <lynx.mw+kde@gmail.com>		* SPDX-FileCopyrightText: 2020 Mathias Wein <lynx.mw+kde@gmail.com>
* SPDX-FileCopyrightText: 2022 L. E. Segovia <amy@amyspark.me>		* SPDX-FileCopyrightText: 2022 L. E. Segovia <amy@amyspark.me>
*		*
* SPDX-License-Identifier: GPL-2.0-or-later		* SPDX-License-Identifier: GPL-2.0-or-later
*/		*/
// for calculation of the needed alignment		// for calculation of the needed alignment
#include <xsimd_extensions/xsimd.hpp>		#include <xsimd_extensions/xsimd.hpp>
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
#include <KoOptimizedCompositeOpOver32.h>		#include <KoOptimizedCompositeOpOver32.h>
#include <KoOptimizedCompositeOpOver128.h>		#include <KoOptimizedCompositeOpOver128.h>
#include <KoOptimizedCompositeOpCopy128.h>		#include <KoOptimizedCompositeOpCopy128.h>
#include <KoOptimizedCompositeOpAlphaDarken32.h>		#include <KoOptimizedCompositeOpAlphaDarken32.h>
#endif		#endif
#include "kis_composition_benchmark.h"		#include "kis_composition_benchmark.h"
#include <simpletest.h>		#include <simpletest.h>
#include <QElapsedTimer>		#include <QElapsedTimer>
skipping to change at line 47		skipping to change at line 47
#include <kis_debug.h>		#include <kis_debug.h>
#if defined Q_OS_WIN		#if defined Q_OS_WIN
#define MEMALIGN_ALLOC(p, a, s) ((*(p)) = _aligned_malloc((s), (a)), *(p) ? 0 : errno)		#define MEMALIGN_ALLOC(p, a, s) ((*(p)) = _aligned_malloc((s), (a)), *(p) ? 0 : errno)
#define MEMALIGN_FREE(p) _aligned_free((p))		#define MEMALIGN_FREE(p) _aligned_free((p))
#else		#else
#define MEMALIGN_ALLOC(p, a, s) posix_memalign((p), (a), (s))		#define MEMALIGN_ALLOC(p, a, s) posix_memalign((p), (a), (s))
#define MEMALIGN_FREE(p) free((p))		#define MEMALIGN_FREE(p) free((p))
#endif		#endif
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
using float_v = xsimd::batch<float, xsimd::current_arch>;		using float_v = xsimd::batch<float, xsimd::current_arch>;
#endif		#endif
enum AlphaRange {		enum AlphaRange {
ALPHA_ZERO,		ALPHA_ZERO,
ALPHA_UNIT,		ALPHA_UNIT,
ALPHA_RANDOM		ALPHA_RANDOM
};		};
template <typename channel_type, class RandomGenerator>		template <typename channel_type, class RandomGenerator>
skipping to change at line 231		skipping to change at line 231
#include <stdint.h>		#include <stdint.h>
QVector<Tile> generateTiles(int size,		QVector<Tile> generateTiles(int size,
const int srcAlignmentShift,		const int srcAlignmentShift,
const int dstAlignmentShift,		const int dstAlignmentShift,
AlphaRange srcAlphaRange,		AlphaRange srcAlphaRange,
AlphaRange dstAlphaRange,		AlphaRange dstAlphaRange,
const quint32 pixelSize)		const quint32 pixelSize)
{		{
QVector<Tile> tiles(size);		QVector<Tile> tiles(size);
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
const int vecSize = float_v::size;		const int vecSize = float_v::size;
#else		#else
const int vecSize = 1;		const int vecSize = 1;
#endif		#endif
// the 256 are used to make sure that we have a good alignment no matter wha t build options are used.		// the 256 are used to make sure that we have a good alignment no matter wha t build options are used.
const size_t pixelAlignment = qMax(size_t(vecSize * sizeof(float)), size_t(2 56));		const size_t pixelAlignment = qMax(size_t(vecSize * sizeof(float)), size_t(2 56));
const size_t maskAlignment = qMax(size_t(vecSize), size_t(256));		const size_t maskAlignment = qMax(size_t(vecSize), size_t(256));
for (int i = 0; i < size; i++) {		for (int i = 0; i < size; i++) {
void *ptr = 0;		void *ptr = 0;
skipping to change at line 505		skipping to change at line 505
benchmarkCompositeOp(op, false, 1.0, 1.0, 0, 0, ALPHA_ZERO, ALPHA_ZERO);		benchmarkCompositeOp(op, false, 1.0, 1.0, 0, 0, ALPHA_ZERO, ALPHA_ZERO);
benchmarkCompositeOp(op, false, 1.0, 1.0, 0, 0, ALPHA_UNIT, ALPHA_ZERO);		benchmarkCompositeOp(op, false, 1.0, 1.0, 0, 0, ALPHA_UNIT, ALPHA_ZERO);
/// ---		/// ---
benchmarkCompositeOp(op, false, 1.0, 1.0, 0, 0, ALPHA_RANDOM, ALPHA_UNIT);		benchmarkCompositeOp(op, false, 1.0, 1.0, 0, 0, ALPHA_RANDOM, ALPHA_UNIT);
benchmarkCompositeOp(op, false, 1.0, 1.0, 0, 0, ALPHA_ZERO, ALPHA_UNIT);		benchmarkCompositeOp(op, false, 1.0, 1.0, 0, 0, ALPHA_ZERO, ALPHA_UNIT);
benchmarkCompositeOp(op, false, 1.0, 1.0, 0, 0, ALPHA_UNIT, ALPHA_UNIT);		benchmarkCompositeOp(op, false, 1.0, 1.0, 0, 0, ALPHA_UNIT, ALPHA_UNIT);
}		}
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
template <typename channels_type>		template <typename channels_type>
void printError(quint8 *s, quint8 *d1, quint8 *d2, quint8 *msk1, int pos)		void printError(quint8 *s, quint8 *d1, quint8 *d2, quint8 *msk1, int pos)
{		{
const channels_type *src1 = reinterpret_cast<const channels_type*>(s);		const channels_type *src1 = reinterpret_cast<const channels_type*>(s);
const channels_type *dst1 = reinterpret_cast<const channels_type*>(d1);		const channels_type *dst1 = reinterpret_cast<const channels_type*>(d1);
const channels_type *dst2 = reinterpret_cast<const channels_type*>(d2);		const channels_type *dst2 = reinterpret_cast<const channels_type*>(d2);
qDebug() << "Wrong rounding in pixel:" << pos;		qDebug() << "Wrong rounding in pixel:" << pos;
qDebug() << "Vector version: " << dst1[0] << dst1[1] << dst1[2] << dst1[3];		qDebug() << "Vector version: " << dst1[0] << dst1[1] << dst1[2] << dst1[3];
skipping to change at line 618		skipping to change at line 618
}		}
}		}
freeTiles(tiles, 0, 0);		freeTiles(tiles, 0, 0);
}		}
#endif		#endif
void KisCompositionBenchmark::detectBuildArchitecture()		void KisCompositionBenchmark::detectBuildArchitecture()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
using namespace xsimd;		using namespace xsimd;
qDebug() << "built for" << ppVar(current_arch().name());		qDebug() << "built for" << ppVar(current_arch().name());
qDebug() << "built for" << ppVar(default_arch().name());		qDebug() << "built for" << ppVar(default_arch().name());
qDebug() << ppVar(supported_architectures().contains<sse2>());		qDebug() << ppVar(supported_architectures().contains<sse2>());
qDebug() << ppVar(supported_architectures().contains<sse3>());		qDebug() << ppVar(supported_architectures().contains<sse3>());
qDebug() << ppVar(supported_architectures().contains<ssse3>());		qDebug() << ppVar(supported_architectures().contains<ssse3>());
qDebug() << ppVar(supported_architectures().contains<sse4_1>());		qDebug() << ppVar(supported_architectures().contains<sse4_1>());
qDebug() << ppVar(supported_architectures().contains<sse4_2>());		qDebug() << ppVar(supported_architectures().contains<sse4_2>());
skipping to change at line 645		skipping to change at line 645
qDebug() << ppVar(supported_architectures().contains<avx512f>());		qDebug() << ppVar(supported_architectures().contains<avx512f>());
qDebug() << ppVar(supported_architectures().contains<avx512bw>());		qDebug() << ppVar(supported_architectures().contains<avx512bw>());
qDebug() << ppVar(supported_architectures().contains<avx512dq>());		qDebug() << ppVar(supported_architectures().contains<avx512dq>());
qDebug() << ppVar(supported_architectures().contains<avx512cd>());		qDebug() << ppVar(supported_architectures().contains<avx512cd>());
qDebug().nospace() << "running on " << hex << "0x" << xsimd::available_archi tectures().best;		qDebug().nospace() << "running on " << hex << "0x" << xsimd::available_archi tectures().best;
#endif		#endif
}		}
void KisCompositionBenchmark::checkRoundingAlphaDarken_05_03()		void KisCompositionBenchmark::checkRoundingAlphaDarken_05_03()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
checkRounding<AlphaDarkenCompositor32<quint8, quint32, KoAlphaDarkenParamsWr apperCreamy> >(0.5,0.3);		checkRounding<AlphaDarkenCompositor32<quint8, quint32, KoAlphaDarkenParamsWr apperCreamy> >(0.5,0.3);
#endif		#endif
}		}
void KisCompositionBenchmark::checkRoundingAlphaDarken_05_05()		void KisCompositionBenchmark::checkRoundingAlphaDarken_05_05()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
checkRounding<AlphaDarkenCompositor32<quint8, quint32, KoAlphaDarkenParamsWr apperCreamy> >(0.5,0.5);		checkRounding<AlphaDarkenCompositor32<quint8, quint32, KoAlphaDarkenParamsWr apperCreamy> >(0.5,0.5);
#endif		#endif
}		}
void KisCompositionBenchmark::checkRoundingAlphaDarken_05_07()		void KisCompositionBenchmark::checkRoundingAlphaDarken_05_07()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
checkRounding<AlphaDarkenCompositor32<quint8, quint32, KoAlphaDarkenParamsWr apperCreamy> >(0.5,0.7);		checkRounding<AlphaDarkenCompositor32<quint8, quint32, KoAlphaDarkenParamsWr apperCreamy> >(0.5,0.7);
#endif		#endif
}		}
void KisCompositionBenchmark::checkRoundingAlphaDarken_05_10()		void KisCompositionBenchmark::checkRoundingAlphaDarken_05_10()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
checkRounding<AlphaDarkenCompositor32<quint8, quint32, KoAlphaDarkenParamsWr apperCreamy> >(0.5,1.0);		checkRounding<AlphaDarkenCompositor32<quint8, quint32, KoAlphaDarkenParamsWr apperCreamy> >(0.5,1.0);
#endif		#endif
}		}
void KisCompositionBenchmark::checkRoundingAlphaDarken_05_10_08()		void KisCompositionBenchmark::checkRoundingAlphaDarken_05_10_08()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
checkRounding<AlphaDarkenCompositor32<quint8, quint32, KoAlphaDarkenParamsWr apperCreamy> >(0.5,1.0,0.8);		checkRounding<AlphaDarkenCompositor32<quint8, quint32, KoAlphaDarkenParamsWr apperCreamy> >(0.5,1.0,0.8);
#endif		#endif
}		}
void KisCompositionBenchmark::checkRoundingAlphaDarkenF32_05_03()		void KisCompositionBenchmark::checkRoundingAlphaDarkenF32_05_03()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
checkRounding<OverCompositor128<float, false, true> >(0.5, 0.3, -1, 16);		checkRounding<OverCompositor128<float, false, true> >(0.5, 0.3, -1, 16);
#endif		#endif
}		}
void KisCompositionBenchmark::checkRoundingAlphaDarkenF32_05_05()		void KisCompositionBenchmark::checkRoundingAlphaDarkenF32_05_05()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
checkRounding<OverCompositor128<float, false, true> >(0.5, 0.5, -1, 16);		checkRounding<OverCompositor128<float, false, true> >(0.5, 0.5, -1, 16);
#endif		#endif
}		}
void KisCompositionBenchmark::checkRoundingAlphaDarkenF32_05_07()		void KisCompositionBenchmark::checkRoundingAlphaDarkenF32_05_07()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
checkRounding<OverCompositor128<float, false, true> >(0.5, 0.7, -1, 16);		checkRounding<OverCompositor128<float, false, true> >(0.5, 0.7, -1, 16);
#endif		#endif
}		}
void KisCompositionBenchmark::checkRoundingAlphaDarkenF32_05_10()		void KisCompositionBenchmark::checkRoundingAlphaDarkenF32_05_10()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
checkRounding<OverCompositor128<float, false, true> >(0.5, 1.0, -1, 16);		checkRounding<OverCompositor128<float, false, true> >(0.5, 1.0, -1, 16);
#endif		#endif
}		}
void KisCompositionBenchmark::checkRoundingAlphaDarkenF32_05_10_08()		void KisCompositionBenchmark::checkRoundingAlphaDarkenF32_05_10_08()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
checkRounding<OverCompositor128<float, false, true> >(0.5, 1.0, 0.8, 16);		checkRounding<OverCompositor128<float, false, true> >(0.5, 1.0, 0.8, 16);
#endif		#endif
}		}
void KisCompositionBenchmark::checkRoundingOver()		void KisCompositionBenchmark::checkRoundingOver()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
checkRounding<OverCompositor32<quint8, quint32, false, true> >(0.5, 0.3);		checkRounding<OverCompositor32<quint8, quint32, false, true> >(0.5, 0.3);
#endif		#endif
}		}
void KisCompositionBenchmark::checkRoundingOverRgbaU16()		void KisCompositionBenchmark::checkRoundingOverRgbaU16()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
checkRounding<OverCompositor128<quint16, false, true> >(0.5, 1.0, -1, 8);		checkRounding<OverCompositor128<quint16, false, true> >(0.5, 1.0, -1, 8);
#endif		#endif
}		}
void KisCompositionBenchmark::checkRoundingOverRgbaF32()		void KisCompositionBenchmark::checkRoundingOverRgbaF32()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
checkRounding<OverCompositor128<float, false, true> >(0.5, 1.0, -1, 16);		checkRounding<OverCompositor128<float, false, true> >(0.5, 1.0, -1, 16);
#endif		#endif
}		}
#include <cfenv>		#include <cfenv>
void KisCompositionBenchmark::checkRoundingCopyRgbaU16()		void KisCompositionBenchmark::checkRoundingCopyRgbaU16()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
checkRounding<CopyCompositor128<quint16, false, true> >(0.5, 1.0, -1, 8);		checkRounding<CopyCompositor128<quint16, false, true> >(0.5, 1.0, -1, 8);
#endif		#endif
}		}
void KisCompositionBenchmark::checkRoundingCopyRgbaF32()		void KisCompositionBenchmark::checkRoundingCopyRgbaF32()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
checkRounding<CopyCompositor128<float, false, true> >(0.5, 1.0, -1, 16);		checkRounding<CopyCompositor128<float, false, true> >(0.5, 1.0, -1, 16);
#endif		#endif
}		}
void KisCompositionBenchmark::compareAlphaDarkenOps()		void KisCompositionBenchmark::compareAlphaDarkenOps()
{		{
const KoColorSpace *cs = KoColorSpaceRegistry::instance()->rgb8();		const KoColorSpace *cs = KoColorSpaceRegistry::instance()->rgb8();
KoCompositeOp *opAct = KoOptimizedCompositeOpFactory::createAlphaDarkenOpCre amy32(cs);		KoCompositeOp *opAct = KoOptimizedCompositeOpFactory::createAlphaDarkenOpCre amy32(cs);
KoCompositeOp *opExp = new KoCompositeOpAlphaDarken<KoBgrU8Traits, KoAlphaDa rkenParamsWrapperCreamy>(cs);		KoCompositeOp *opExp = new KoCompositeOpAlphaDarken<KoBgrU8Traits, KoAlphaDa rkenParamsWrapperCreamy>(cs);
skipping to change at line 1055		skipping to change at line 1055
QBENCHMARK_ONCE {		QBENCHMARK_ONCE {
Q_FOREACH (const Tile &tile, tiles) {		Q_FOREACH (const Tile &tile, tiles) {
memcpy(tile.dst, tile.src, 4 * numPixels);		memcpy(tile.dst, tile.src, 4 * numPixels);
}		}
}		}
freeTiles(tiles, 0, 0);		freeTiles(tiles, 0, 0);
}		}
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN
const int vecSize = float_v::size;		IVERSAL_BUILD_PASS
const size_t uint8VecAlignment = qMax(vecSize * sizeof(quint8), sizeof(void*		const int vecSize = float_v::size;
));		const size_t uint8VecAlignment = qMax(vecSize * sizeof(quint8), sizeof(void *));
const size_t uint32VecAlignment = qMax(vecSize * sizeof(quint32), sizeof(voi		const size_t uint32VecAlignment = qMax(vecSize * sizeof(quint32), sizeof(void *)
d*));		);
const size_t floatVecAlignment = qMax(vecSize * sizeof(float), sizeof(void*)		const size_t floatVecAlignment = qMax(vecSize * sizeof(float), sizeof(void *));
);
#endif		#endif
void KisCompositionBenchmark::benchmarkUintFloat()		void KisCompositionBenchmark::benchmarkUintFloat()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
using uint_v = xsimd::batch<unsigned int, xsimd::current_arch>;		using uint_v = xsimd::batch<unsigned int, xsimd::current_arch>;
const int dataSize = 4096;		const int dataSize = 4096;
void *ptr = 0;		void *ptr = 0;
int error = MEMALIGN_ALLOC(&ptr, uint8VecAlignment, dataSize);		int error = MEMALIGN_ALLOC(&ptr, uint8VecAlignment, dataSize);
if (error) {		if (error) {
qFatal("posix_memalign failed: %d", error);		qFatal("posix_memalign failed: %d", error);
}		}
quint8 *iData = (quint8*)ptr;		quint8 *iData = (quint8*)ptr;
error = MEMALIGN_ALLOC(&ptr, floatVecAlignment, dataSize * sizeof(float));		error = MEMALIGN_ALLOC(&ptr, floatVecAlignment, dataSize * sizeof(float));
skipping to change at line 1098		skipping to change at line 1098
}		}
}		}
MEMALIGN_FREE(iData);		MEMALIGN_FREE(iData);
MEMALIGN_FREE(fData);		MEMALIGN_FREE(fData);
#endif		#endif
}		}
void KisCompositionBenchmark::benchmarkUintIntFloat()		void KisCompositionBenchmark::benchmarkUintIntFloat()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
using uint_v = xsimd::batch<unsigned int, xsimd::current_arch>;		using uint_v = xsimd::batch<unsigned int, xsimd::current_arch>;
const int dataSize = 4096;		const int dataSize = 4096;
void *ptr = 0;		void *ptr = 0;
int error = MEMALIGN_ALLOC(&ptr, uint8VecAlignment, dataSize);		int error = MEMALIGN_ALLOC(&ptr, uint8VecAlignment, dataSize);
if (error) {		if (error) {
qFatal("posix_memalign failed: %d", error);		qFatal("posix_memalign failed: %d", error);
}		}
quint8 *iData = (quint8*)ptr;		quint8 *iData = (quint8*)ptr;
error = MEMALIGN_ALLOC(&ptr, floatVecAlignment, dataSize * sizeof(float));		error = MEMALIGN_ALLOC(&ptr, floatVecAlignment, dataSize * sizeof(float));
skipping to change at line 1130		skipping to change at line 1130
}		}
}		}
MEMALIGN_FREE(iData);		MEMALIGN_FREE(iData);
MEMALIGN_FREE(fData);		MEMALIGN_FREE(fData);
#endif		#endif
}		}
void KisCompositionBenchmark::benchmarkFloatUint()		void KisCompositionBenchmark::benchmarkFloatUint()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
using uint_v = xsimd::batch<unsigned int, xsimd::current_arch>;		using uint_v = xsimd::batch<unsigned int, xsimd::current_arch>;
const int dataSize = 4096;		const int dataSize = 4096;
void *ptr = 0;		void *ptr = 0;
int error = MEMALIGN_ALLOC(&ptr, uint32VecAlignment, dataSize * sizeof(quint 32));		int error = MEMALIGN_ALLOC(&ptr, uint32VecAlignment, dataSize * sizeof(quint 32));
if (error) {		if (error) {
qFatal("posix_memalign failed: %d", error);		qFatal("posix_memalign failed: %d", error);
}		}
quint32 *iData = (quint32*)ptr;		quint32 *iData = (quint32*)ptr;
error = MEMALIGN_ALLOC(&ptr, floatVecAlignment, dataSize * sizeof(float));		error = MEMALIGN_ALLOC(&ptr, floatVecAlignment, dataSize * sizeof(float));
skipping to change at line 1163		skipping to change at line 1163
}		}
}		}
MEMALIGN_FREE(iData);		MEMALIGN_FREE(iData);
MEMALIGN_FREE(fData);		MEMALIGN_FREE(fData);
#endif		#endif
}		}
void KisCompositionBenchmark::benchmarkFloatIntUint()		void KisCompositionBenchmark::benchmarkFloatIntUint()
{		{
#if defined HAVE_XSIMD && XSIMD_UNIVERSAL_BUILD_PASS		#if defined(HAVE_XSIMD) && !defined(XSIMD_NO_SUPPORTED_ARCHITECTURE) && XSIMD_UN IVERSAL_BUILD_PASS
using uint_v = xsimd::batch<unsigned int, xsimd::current_arch>;		using uint_v = xsimd::batch<unsigned int, xsimd::current_arch>;
const int dataSize = 4096;		const int dataSize = 4096;
void *ptr = 0;		void *ptr = 0;
int error = MEMALIGN_ALLOC(&ptr, uint32VecAlignment, dataSize * sizeof(quint 32));		int error = MEMALIGN_ALLOC(&ptr, uint32VecAlignment, dataSize * sizeof(quint 32));
if (error) {		if (error) {
qFatal("posix_memalign failed: %d", error);		qFatal("posix_memalign failed: %d", error);
}		}
quint32 *iData = (quint32*)ptr;		quint32 *iData = (quint32*)ptr;
error = MEMALIGN_ALLOC(&ptr, floatVecAlignment, dataSize * sizeof(float));		error = MEMALIGN_ALLOC(&ptr, floatVecAlignment, dataSize * sizeof(float));
if (error) {		if (error) {
End of changes. 25 change blocks.
32 lines changed or deleted		31 lines changed or added

---