1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
|
// Copyright (C) 2016 The Qt Company Ltd.
// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only
#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>
#include <private/qthreadpool_p.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 = QThreadPoolPrivate::qtGuiInstance();
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
|
