/src/qtbase/src/gui/painting/qimagescale_sse4.cpp

Source (jump to first uncovered line)
/****************************************************************************
**
** Copyright (C) 2016 The Qt Company Ltd.
** Contact: https://www.qt.io/licensing/
**
** This file is part of the QtGui module of the Qt Toolkit.
**
** $QT_BEGIN_LICENSE:LGPL$
** Commercial License Usage
** Licensees holding valid commercial Qt licenses may use this file in
** accordance with the commercial license agreement provided with the
** Software or, alternatively, in accordance with the terms contained in
** a written agreement between you and The Qt Company. For licensing terms
** and conditions see https://www.qt.io/terms-conditions. For further
** information use the contact form at https://www.qt.io/contact-us.
**
** GNU Lesser General Public License Usage
** Alternatively, this file may be used under the terms of the GNU Lesser
** General Public License version 3 as published by the Free Software
** Foundation and appearing in the file LICENSE.LGPL3 included in the
** packaging of this file. Please review the following information to
** ensure the GNU Lesser General Public License version 3 requirements
** will be met: https://www.gnu.org/licenses/lgpl-3.0.html.
**
** GNU General Public License Usage
** Alternatively, this file may be used under the terms of the GNU
** General Public License version 2.0 or (at your option) the GNU General
** Public license version 3 or any later version approved by the KDE Free
** Qt Foundation. The licenses are as published by the Free Software
** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3
** included in the packaging of this file. Please review the following
** information to ensure the GNU General Public License requirements will
** be met: https://www.gnu.org/licenses/gpl-2.0.html and
** https://www.gnu.org/licenses/gpl-3.0.html.
**
** $QT_END_LICENSE$
**
****************************************************************************/

#include "qimagescale_p.h"
#include "qimage.h"
#include <private/qdrawhelper_x86_p.h>
#include <private/qsimd_p.h>

#if QT_CONFIG(thread) && !defined(Q_OS_WASM)
#include "qsemaphore.h"
#include "qthreadpool.h"
#endif

#if defined(QT_COMPILER_SUPPORTS_SSE4_1)

QT_BEGIN_NAMESPACE

using namespace QImageScale;

template<typename T>
static inline void multithread_pixels_function(QImageScaleInfo *isi, int dh, const T &scaleSection)
{
#if QT_CONFIG(thread) && !defined(Q_OS_WASM)
    int segments = (qsizetype(isi->sh) * isi->sw) / (1<<16);
    segments = std::min(segments, dh);
    QThreadPool *threadPool = QThreadPool::globalInstance();
    if (segments > 1 && threadPool && !threadPool->contains(QThread::currentThread())) {
        QSemaphore semaphore;
        int y = 0;
        for (int i = 0; i < segments; ++i) {
            int yn = (dh - y) / (segments - i);
            threadPool->start([&, y, yn]() {
                scaleSection(y, y + yn);
                semaphore.release(1);
            });
            y += yn;
        }
        semaphore.acquire(segments);
        return;
    }
#endif
    scaleSection(0, dh);
}

inline static __m128i Q_DECL_VECTORCALL
qt_qimageScaleAARGBA_helper(const unsigned int *pix, int xyap, int Cxy, int step, const __m128i vxyap, const __m128i vCxy)
{
    __m128i vpix = _mm_cvtepu8_epi32(_mm_cvtsi32_si128(*pix));
    __m128i vx = _mm_mullo_epi32(vpix, vxyap);
    int i;
    for (i = (1 << 14) - xyap; i > Cxy; i -= Cxy) {
        pix += step;
        vpix = _mm_cvtepu8_epi32(_mm_cvtsi32_si128(*pix));
        vx = _mm_add_epi32(vx, _mm_mullo_epi32(vpix, vCxy));
    }
    pix += step;
    vpix = _mm_cvtepu8_epi32(_mm_cvtsi32_si128(*pix));
    vx = _mm_add_epi32(vx, _mm_mullo_epi32(vpix, _mm_set1_epi32(i)));
    return vx;
}

template<bool RGB>
void qt_qimageScaleAARGBA_up_x_down_y_sse4(QImageScaleInfo *isi, unsigned int *dest,
                                           int dw, int dh, int dow, int sow)
{
    const unsigned int **ypoints = isi->ypoints;
    const int *xpoints = isi->xpoints;
    const int *xapoints = isi->xapoints;
    const int *yapoints = isi->yapoints;

    const __m128i v256 = _mm_set1_epi32(256);

    /* go through every scanline in the output buffer */
    auto scaleSection = [&] (int yStart, int yEnd) {
        for (int y = yStart; y < yEnd; ++y) {
            const int Cy = yapoints[y] >> 16;
            const int yap = yapoints[y] & 0xffff;
            const __m128i vCy = _mm_set1_epi32(Cy);
            const __m128i vyap = _mm_set1_epi32(yap);

            unsigned int *dptr = dest + (y * dow);
            for (int x = 0; x < dw; x++) {
                const unsigned int *sptr = ypoints[y] + xpoints[x];
                __m128i vx = qt_qimageScaleAARGBA_helper(sptr, yap, Cy, sow, vyap, vCy);

                const int xap = xapoints[x];
                if (xap > 0) {
                    const __m128i vxap = _mm_set1_epi32(xap);
                    const __m128i vinvxap = _mm_sub_epi32(v256, vxap);
                    __m128i vr = qt_qimageScaleAARGBA_helper(sptr + 1, yap, Cy, sow, vyap, vCy);

                    vx = _mm_mullo_epi32(vx, vinvxap);
                    vr = _mm_mullo_epi32(vr, vxap);
                    vx = _mm_add_epi32(vx, vr);
                    vx = _mm_srli_epi32(vx, 8);
                }
                vx = _mm_srli_epi32(vx, 14);
                vx = _mm_packus_epi32(vx, vx);
                vx = _mm_packus_epi16(vx, vx);
                *dptr = _mm_cvtsi128_si32(vx);
                if (RGB)
                    *dptr |= 0xff000000;
                dptr++;
            }
        }
    };
    multithread_pixels_function(isi, dh, scaleSection);
}

template<bool RGB>
void qt_qimageScaleAARGBA_down_x_up_y_sse4(QImageScaleInfo *isi, unsigned int *dest,
                                           int dw, int dh, int dow, int sow)
{
    const unsigned int **ypoints = isi->ypoints;
    int *xpoints = isi->xpoints;
    int *xapoints = isi->xapoints;
    int *yapoints = isi->yapoints;

    const __m128i v256 = _mm_set1_epi32(256);

    /* go through every scanline in the output buffer */
    auto scaleSection = [&] (int yStart, int yEnd) {
        for (int y = yStart; y < yEnd; ++y) {
            unsigned int *dptr = dest + (y * dow);
            for (int x = 0; x < dw; x++) {
                int Cx = xapoints[x] >> 16;
                int xap = xapoints[x] & 0xffff;
                const __m128i vCx = _mm_set1_epi32(Cx);
                const __m128i vxap = _mm_set1_epi32(xap);

                const unsigned int *sptr = ypoints[y] + xpoints[x];
                __m128i vx = qt_qimageScaleAARGBA_helper(sptr, xap, Cx, 1, vxap, vCx);

                int yap = yapoints[y];
                if (yap > 0) {
                    const __m128i vyap = _mm_set1_epi32(yap);
                    const __m128i vinvyap = _mm_sub_epi32(v256, vyap);
                    __m128i vr = qt_qimageScaleAARGBA_helper(sptr + sow, xap, Cx, 1, vxap, vCx);

                    vx = _mm_mullo_epi32(vx, vinvyap);
                    vr = _mm_mullo_epi32(vr, vyap);
                    vx = _mm_add_epi32(vx, vr);
                    vx = _mm_srli_epi32(vx, 8);
                }
                vx = _mm_srli_epi32(vx, 14);
                vx = _mm_packus_epi32(vx, vx);
                vx = _mm_packus_epi16(vx, vx);
                *dptr = _mm_cvtsi128_si32(vx);
                if (RGB)
                    *dptr |= 0xff000000;
                dptr++;
            }
        }
    };
    multithread_pixels_function(isi, dh, scaleSection);
}

template<bool RGB>
void qt_qimageScaleAARGBA_down_xy_sse4(QImageScaleInfo *isi, unsigned int *dest,
                                       int dw, int dh, int dow, int sow)
{
    const unsigned int **ypoints = isi->ypoints;
    int *xpoints = isi->xpoints;
    int *xapoints = isi->xapoints;
    int *yapoints = isi->yapoints;

    auto scaleSection = [&] (int yStart, int yEnd) {
        for (int y = yStart; y < yEnd; ++y) {
            int Cy = yapoints[y] >> 16;
            int yap = yapoints[y] & 0xffff;
            const __m128i vCy = _mm_set1_epi32(Cy);
            const __m128i vyap = _mm_set1_epi32(yap);

            unsigned int *dptr = dest + (y * dow);
            for (int x = 0; x < dw; x++) {
                const int Cx = xapoints[x] >> 16;
                const int xap = xapoints[x] & 0xffff;
                const __m128i vCx = _mm_set1_epi32(Cx);
                const __m128i vxap = _mm_set1_epi32(xap);

                const unsigned int *sptr = ypoints[y] + xpoints[x];
                __m128i vx = qt_qimageScaleAARGBA_helper(sptr, xap, Cx, 1, vxap, vCx);
                __m128i vr = _mm_mullo_epi32(_mm_srli_epi32(vx, 4), vyap);

                int j;
                for (j = (1 << 14) - yap; j > Cy; j -= Cy) {
                    sptr += sow;
                    vx = qt_qimageScaleAARGBA_helper(sptr, xap, Cx, 1, vxap, vCx);
                    vr = _mm_add_epi32(vr, _mm_mullo_epi32(_mm_srli_epi32(vx, 4), vCy));
                }
                sptr += sow;
                vx = qt_qimageScaleAARGBA_helper(sptr, xap, Cx, 1, vxap, vCx);
                vr = _mm_add_epi32(vr, _mm_mullo_epi32(_mm_srli_epi32(vx, 4), _mm_set1_epi32(j)));

                vr = _mm_srli_epi32(vr, 24);
                vr = _mm_packus_epi32(vr, _mm_setzero_si128());
                vr = _mm_packus_epi16(vr, _mm_setzero_si128());
                *dptr = _mm_cvtsi128_si32(vr);
                if (RGB)
                    *dptr |= 0xff000000;
                dptr++;
            }
        }
    };
    multithread_pixels_function(isi, dh, scaleSection);
}

template void qt_qimageScaleAARGBA_up_x_down_y_sse4<false>(QImageScaleInfo *isi, unsigned int *dest,
                                                           int dw, int dh, int dow, int sow);

template void qt_qimageScaleAARGBA_up_x_down_y_sse4<true>(QImageScaleInfo *isi, unsigned int *dest,
                                                          int dw, int dh, int dow, int sow);

template void qt_qimageScaleAARGBA_down_x_up_y_sse4<false>(QImageScaleInfo *isi, unsigned int *dest,
                                                           int dw, int dh, int dow, int sow);

template void qt_qimageScaleAARGBA_down_x_up_y_sse4<true>(QImageScaleInfo *isi, unsigned int *dest,
                                                          int dw, int dh, int dow, int sow);

template void qt_qimageScaleAARGBA_down_xy_sse4<false>(QImageScaleInfo *isi, unsigned int *dest,
                                                       int dw, int dh, int dow, int sow);

template void qt_qimageScaleAARGBA_down_xy_sse4<true>(QImageScaleInfo *isi, unsigned int *dest,
                                                      int dw, int dh, int dow, int sow);

QT_END_NAMESPACE

#endif

Coverage Report

Created: 2025-07-12 07:23

Line	Count	Source (jump to first uncovered line)
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		#include "qimagescale_p.h"
41		#include "qimage.h"
42		#include <private/qdrawhelper_x86_p.h>
43		#include <private/qsimd_p.h>
44
45		#if QT_CONFIG(thread) && !defined(Q_OS_WASM)
46		#include "qsemaphore.h"
47		#include "qthreadpool.h"
48		#endif
49
50		#if defined(QT_COMPILER_SUPPORTS_SSE4_1)
51
52		QT_BEGIN_NAMESPACE
53
54		using namespace QImageScale;
55
56		template<typename T>
57		static inline void multithread_pixels_function(QImageScaleInfo *isi, int dh, const T &scaleSection)
58	0	{
59	0	#if QT_CONFIG(thread) && !defined(Q_OS_WASM)
60	0	int segments = (qsizetype(isi->sh) * isi->sw) / (1<<16);
61	0	segments = std::min(segments, dh);
62	0	QThreadPool *threadPool = QThreadPool::globalInstance();
63	0	if (segments > 1 && threadPool && !threadPool->contains(QThread::currentThread())) {
64	0	QSemaphore semaphore;
65	0	int y = 0;
66	0	for (int i = 0; i < segments; ++i) {
67	0	int yn = (dh - y) / (segments - i);
68	0	threadPool->start([&, y, yn]() {
69	0	scaleSection(y, y + yn);
70	0	semaphore.release(1);
71	0	}); Unexecuted instantiation: qimagescale_sse4.cpp:multithread_pixels_function<qt_qimageScaleAARGBA_up_x_down_y_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}>(QImageScale::QImageScaleInfo, int, qt_qimageScaleAARGBA_up_x_down_y_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1} const&)::{lambda()#1}::operator()() const Unexecuted instantiation: qimagescale_sse4.cpp:multithread_pixels_function<qt_qimageScaleAARGBA_up_x_down_y_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}>(QImageScale::QImageScaleInfo, int, qt_qimageScaleAARGBA_up_x_down_y_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1} const&)::{lambda()#1}::operator()() const Unexecuted instantiation: qimagescale_sse4.cpp:multithread_pixels_function<qt_qimageScaleAARGBA_down_x_up_y_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}>(QImageScale::QImageScaleInfo, int, qt_qimageScaleAARGBA_down_x_up_y_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1} const&)::{lambda()#1}::operator()() const Unexecuted instantiation: qimagescale_sse4.cpp:multithread_pixels_function<qt_qimageScaleAARGBA_down_x_up_y_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}>(QImageScale::QImageScaleInfo, int, qt_qimageScaleAARGBA_down_x_up_y_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1} const&)::{lambda()#1}::operator()() const Unexecuted instantiation: qimagescale_sse4.cpp:multithread_pixels_function<qt_qimageScaleAARGBA_down_xy_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}>(QImageScale::QImageScaleInfo, int, qt_qimageScaleAARGBA_down_xy_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1} const&)::{lambda()#1}::operator()() const Unexecuted instantiation: qimagescale_sse4.cpp:multithread_pixels_function<qt_qimageScaleAARGBA_down_xy_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}>(QImageScale::QImageScaleInfo, int, qt_qimageScaleAARGBA_down_xy_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1} const&)::{lambda()#1}::operator()() const
72	0	y += yn;
73	0	}
74	0	semaphore.acquire(segments);
75	0	return;
76	0	}
77	0	#endif
78	0	scaleSection(0, dh);
79	0	} Unexecuted instantiation: qimagescale_sse4.cpp:void multithread_pixels_function<qt_qimageScaleAARGBA_up_x_down_y_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}>(QImageScale::QImageScaleInfo, int, qt_qimageScaleAARGBA_up_x_down_y_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1} const&) Unexecuted instantiation: qimagescale_sse4.cpp:void multithread_pixels_function<qt_qimageScaleAARGBA_up_x_down_y_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}>(QImageScale::QImageScaleInfo, int, qt_qimageScaleAARGBA_up_x_down_y_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1} const&) Unexecuted instantiation: qimagescale_sse4.cpp:void multithread_pixels_function<qt_qimageScaleAARGBA_down_x_up_y_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}>(QImageScale::QImageScaleInfo, int, qt_qimageScaleAARGBA_down_x_up_y_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1} const&) Unexecuted instantiation: qimagescale_sse4.cpp:void multithread_pixels_function<qt_qimageScaleAARGBA_down_x_up_y_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}>(QImageScale::QImageScaleInfo, int, qt_qimageScaleAARGBA_down_x_up_y_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1} const&) Unexecuted instantiation: qimagescale_sse4.cpp:void multithread_pixels_function<qt_qimageScaleAARGBA_down_xy_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}>(QImageScale::QImageScaleInfo, int, qt_qimageScaleAARGBA_down_xy_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1} const&) Unexecuted instantiation: qimagescale_sse4.cpp:void multithread_pixels_function<qt_qimageScaleAARGBA_down_xy_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}>(QImageScale::QImageScaleInfo, int, qt_qimageScaleAARGBA_down_xy_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1} const&)
80
81		inline static __m128i Q_DECL_VECTORCALL
82		qt_qimageScaleAARGBA_helper(const unsigned int *pix, int xyap, int Cxy, int step, const __m128i vxyap, const __m128i vCxy)
83	0	{
84	0	__m128i vpix = _mm_cvtepu8_epi32(_mm_cvtsi32_si128(*pix));
85	0	__m128i vx = _mm_mullo_epi32(vpix, vxyap);
86	0	int i;
87	0	for (i = (1 << 14) - xyap; i > Cxy; i -= Cxy) {
88	0	pix += step;
89	0	vpix = _mm_cvtepu8_epi32(_mm_cvtsi32_si128(*pix));
90	0	vx = _mm_add_epi32(vx, _mm_mullo_epi32(vpix, vCxy));
91	0	}
92	0	pix += step;
93	0	vpix = _mm_cvtepu8_epi32(_mm_cvtsi32_si128(*pix));
94	0	vx = _mm_add_epi32(vx, _mm_mullo_epi32(vpix, _mm_set1_epi32(i)));
95	0	return vx;
96	0	}
97
98		template<bool RGB>
99		void qt_qimageScaleAARGBA_up_x_down_y_sse4(QImageScaleInfo isi, unsigned int dest,
100		int dw, int dh, int dow, int sow)
101	0	{
102	0	const unsigned int **ypoints = isi->ypoints;
103	0	const int *xpoints = isi->xpoints;
104	0	const int *xapoints = isi->xapoints;
105	0	const int *yapoints = isi->yapoints;
106
107	0	const __m128i v256 = _mm_set1_epi32(256);
108
109		/* go through every scanline in the output buffer */
110	0	auto scaleSection = [&] (int yStart, int yEnd) {
111	0	for (int y = yStart; y < yEnd; ++y) {
112	0	const int Cy = yapoints[y] >> 16;
113	0	const int yap = yapoints[y] & 0xffff;
114	0	const __m128i vCy = _mm_set1_epi32(Cy);
115	0	const __m128i vyap = _mm_set1_epi32(yap);
116
117	0	unsigned int dptr = dest + (y dow);
118	0	for (int x = 0; x < dw; x++) {
119	0	const unsigned int *sptr = ypoints[y] + xpoints[x];
120	0	__m128i vx = qt_qimageScaleAARGBA_helper(sptr, yap, Cy, sow, vyap, vCy);
121
122	0	const int xap = xapoints[x];
123	0	if (xap > 0) {
124	0	const __m128i vxap = _mm_set1_epi32(xap);
125	0	const __m128i vinvxap = _mm_sub_epi32(v256, vxap);
126	0	__m128i vr = qt_qimageScaleAARGBA_helper(sptr + 1, yap, Cy, sow, vyap, vCy);
127
128	0	vx = _mm_mullo_epi32(vx, vinvxap);
129	0	vr = _mm_mullo_epi32(vr, vxap);
130	0	vx = _mm_add_epi32(vx, vr);
131	0	vx = _mm_srli_epi32(vx, 8);
132	0	}
133	0	vx = _mm_srli_epi32(vx, 14);
134	0	vx = _mm_packus_epi32(vx, vx);
135	0	vx = _mm_packus_epi16(vx, vx);
136	0	*dptr = _mm_cvtsi128_si32(vx);
137	0	if (RGB)
138	0	*dptr \|= 0xff000000;
139	0	dptr++;
140	0	}
141	0	}
142	0	}; Unexecuted instantiation: qt_qimageScaleAARGBA_up_x_down_y_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}::operator()(int, int) const Unexecuted instantiation: qt_qimageScaleAARGBA_up_x_down_y_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}::operator()(int, int) const
143	0	multithread_pixels_function(isi, dh, scaleSection);
144	0	} Unexecuted instantiation: void qt_qimageScaleAARGBA_up_x_down_y_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int) Unexecuted instantiation: void qt_qimageScaleAARGBA_up_x_down_y_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)
145
146		template<bool RGB>
147		void qt_qimageScaleAARGBA_down_x_up_y_sse4(QImageScaleInfo isi, unsigned int dest,
148		int dw, int dh, int dow, int sow)
149	0	{
150	0	const unsigned int **ypoints = isi->ypoints;
151	0	int *xpoints = isi->xpoints;
152	0	int *xapoints = isi->xapoints;
153	0	int *yapoints = isi->yapoints;
154
155	0	const __m128i v256 = _mm_set1_epi32(256);
156
157		/* go through every scanline in the output buffer */
158	0	auto scaleSection = [&] (int yStart, int yEnd) {
159	0	for (int y = yStart; y < yEnd; ++y) {
160	0	unsigned int dptr = dest + (y dow);
161	0	for (int x = 0; x < dw; x++) {
162	0	int Cx = xapoints[x] >> 16;
163	0	int xap = xapoints[x] & 0xffff;
164	0	const __m128i vCx = _mm_set1_epi32(Cx);
165	0	const __m128i vxap = _mm_set1_epi32(xap);
166
167	0	const unsigned int *sptr = ypoints[y] + xpoints[x];
168	0	__m128i vx = qt_qimageScaleAARGBA_helper(sptr, xap, Cx, 1, vxap, vCx);
169
170	0	int yap = yapoints[y];
171	0	if (yap > 0) {
172	0	const __m128i vyap = _mm_set1_epi32(yap);
173	0	const __m128i vinvyap = _mm_sub_epi32(v256, vyap);
174	0	__m128i vr = qt_qimageScaleAARGBA_helper(sptr + sow, xap, Cx, 1, vxap, vCx);
175
176	0	vx = _mm_mullo_epi32(vx, vinvyap);
177	0	vr = _mm_mullo_epi32(vr, vyap);
178	0	vx = _mm_add_epi32(vx, vr);
179	0	vx = _mm_srli_epi32(vx, 8);
180	0	}
181	0	vx = _mm_srli_epi32(vx, 14);
182	0	vx = _mm_packus_epi32(vx, vx);
183	0	vx = _mm_packus_epi16(vx, vx);
184	0	*dptr = _mm_cvtsi128_si32(vx);
185	0	if (RGB)
186	0	*dptr \|= 0xff000000;
187	0	dptr++;
188	0	}
189	0	}
190	0	}; Unexecuted instantiation: qt_qimageScaleAARGBA_down_x_up_y_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}::operator()(int, int) const Unexecuted instantiation: qt_qimageScaleAARGBA_down_x_up_y_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}::operator()(int, int) const
191	0	multithread_pixels_function(isi, dh, scaleSection);
192	0	} Unexecuted instantiation: void qt_qimageScaleAARGBA_down_x_up_y_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int) Unexecuted instantiation: void qt_qimageScaleAARGBA_down_x_up_y_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)
193
194		template<bool RGB>
195		void qt_qimageScaleAARGBA_down_xy_sse4(QImageScaleInfo isi, unsigned int dest,
196		int dw, int dh, int dow, int sow)
197	0	{
198	0	const unsigned int **ypoints = isi->ypoints;
199	0	int *xpoints = isi->xpoints;
200	0	int *xapoints = isi->xapoints;
201	0	int *yapoints = isi->yapoints;
202
203	0	auto scaleSection = [&] (int yStart, int yEnd) {
204	0	for (int y = yStart; y < yEnd; ++y) {
205	0	int Cy = yapoints[y] >> 16;
206	0	int yap = yapoints[y] & 0xffff;
207	0	const __m128i vCy = _mm_set1_epi32(Cy);
208	0	const __m128i vyap = _mm_set1_epi32(yap);
209
210	0	unsigned int dptr = dest + (y dow);
211	0	for (int x = 0; x < dw; x++) {
212	0	const int Cx = xapoints[x] >> 16;
213	0	const int xap = xapoints[x] & 0xffff;
214	0	const __m128i vCx = _mm_set1_epi32(Cx);
215	0	const __m128i vxap = _mm_set1_epi32(xap);
216
217	0	const unsigned int *sptr = ypoints[y] + xpoints[x];
218	0	__m128i vx = qt_qimageScaleAARGBA_helper(sptr, xap, Cx, 1, vxap, vCx);
219	0	__m128i vr = _mm_mullo_epi32(_mm_srli_epi32(vx, 4), vyap);
220
221	0	int j;
222	0	for (j = (1 << 14) - yap; j > Cy; j -= Cy) {
223	0	sptr += sow;
224	0	vx = qt_qimageScaleAARGBA_helper(sptr, xap, Cx, 1, vxap, vCx);
225	0	vr = _mm_add_epi32(vr, _mm_mullo_epi32(_mm_srli_epi32(vx, 4), vCy));
226	0	}
227	0	sptr += sow;
228	0	vx = qt_qimageScaleAARGBA_helper(sptr, xap, Cx, 1, vxap, vCx);
229	0	vr = _mm_add_epi32(vr, _mm_mullo_epi32(_mm_srli_epi32(vx, 4), _mm_set1_epi32(j)));
230
231	0	vr = _mm_srli_epi32(vr, 24);
232	0	vr = _mm_packus_epi32(vr, _mm_setzero_si128());
233	0	vr = _mm_packus_epi16(vr, _mm_setzero_si128());
234	0	*dptr = _mm_cvtsi128_si32(vr);
235	0	if (RGB)
236	0	*dptr \|= 0xff000000;
237	0	dptr++;
238	0	}
239	0	}
240	0	}; Unexecuted instantiation: qt_qimageScaleAARGBA_down_xy_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}::operator()(int, int) const Unexecuted instantiation: qt_qimageScaleAARGBA_down_xy_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)::{lambda(int, int)#1}::operator()(int, int) const
241	0	multithread_pixels_function(isi, dh, scaleSection);
242	0	} Unexecuted instantiation: void qt_qimageScaleAARGBA_down_xy_sse4<false>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int) Unexecuted instantiation: void qt_qimageScaleAARGBA_down_xy_sse4<true>(QImageScale::QImageScaleInfo, unsigned int, int, int, int, int)
243
244		template void qt_qimageScaleAARGBA_up_x_down_y_sse4<false>(QImageScaleInfo isi, unsigned int dest,
245		int dw, int dh, int dow, int sow);
246
247		template void qt_qimageScaleAARGBA_up_x_down_y_sse4<true>(QImageScaleInfo isi, unsigned int dest,
248		int dw, int dh, int dow, int sow);
249
250		template void qt_qimageScaleAARGBA_down_x_up_y_sse4<false>(QImageScaleInfo isi, unsigned int dest,
251		int dw, int dh, int dow, int sow);
252
253		template void qt_qimageScaleAARGBA_down_x_up_y_sse4<true>(QImageScaleInfo isi, unsigned int dest,
254		int dw, int dh, int dow, int sow);
255
256		template void qt_qimageScaleAARGBA_down_xy_sse4<false>(QImageScaleInfo isi, unsigned int dest,
257		int dw, int dh, int dow, int sow);
258
259		template void qt_qimageScaleAARGBA_down_xy_sse4<true>(QImageScaleInfo isi, unsigned int dest,
260		int dw, int dh, int dow, int sow);
261
262		QT_END_NAMESPACE
263
264		#endif