This repository has been archived by the owner on Aug 4, 2022. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 5
/
Copy pathwin64.patch
210 lines (210 loc) · 7.95 KB
/
win64.patch
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
diff --git a/gfx/ycbcr/yuv_row_win64.cpp b/gfx/ycbcr/yuv_row_win64.cpp
new file mode 100644
--- /dev/null
+++ b/gfx/ycbcr/yuv_row_win64.cpp
@@ -0,0 +1,205 @@
+// Copyright (c) 2010 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "yuv_row.h"
+
+extern "C" {
+
+// x64 compiler doesn't support MMX and inline assembler. Use SSE2 intrinsics.
+
+#define kCoefficientsRgbU (reinterpret_cast<uint8*>(kCoefficientsRgbY) + 2048)
+#define kCoefficientsRgbV (reinterpret_cast<uint8*>(kCoefficientsRgbY) + 4096)
+
+#include <emmintrin.h>
+
+static void FastConvertYUVToRGB32Row_SSE2(const uint8* y_buf,
+ const uint8* u_buf,
+ const uint8* v_buf,
+ uint8* rgb_buf,
+ int width) {
+ __m128i xmm0, xmmY1, xmmY2;
+ __m128 xmmY;
+
+ while (width >= 2) {
+ xmm0 = _mm_adds_epi16(_mm_loadl_epi64(reinterpret_cast<__m128i*>(kCoefficientsRgbU + 8 * *u_buf++)),
+ _mm_loadl_epi64(reinterpret_cast<__m128i*>(kCoefficientsRgbV + 8 * *v_buf++)));
+
+ xmmY1 = _mm_loadl_epi64(reinterpret_cast<__m128i*>(reinterpret_cast<uint8*>(kCoefficientsRgbY) + 8 * *y_buf++));
+ xmmY1 = _mm_adds_epi16(xmmY1, xmm0);
+
+ xmmY2 = _mm_loadl_epi64(reinterpret_cast<__m128i*>(reinterpret_cast<uint8*>(kCoefficientsRgbY) + 8 * *y_buf++));
+ xmmY2 = _mm_adds_epi16(xmmY2, xmm0);
+
+ xmmY = _mm_shuffle_ps(_mm_castsi128_ps(xmmY1), _mm_castsi128_ps(xmmY2),
+ 0x44);
+ xmmY1 = _mm_srai_epi16(_mm_castps_si128(xmmY), 6);
+ xmmY1 = _mm_packus_epi16(xmmY1, xmmY1);
+
+ _mm_storel_epi64(reinterpret_cast<__m128i*>(rgb_buf), xmmY1);
+ rgb_buf += 8;
+ width -= 2;
+ }
+
+ if (width) {
+ xmm0 = _mm_adds_epi16(_mm_loadl_epi64(reinterpret_cast<__m128i*>(kCoefficientsRgbU + 8 * *u_buf)),
+ _mm_loadl_epi64(reinterpret_cast<__m128i*>(kCoefficientsRgbV + 8 * *v_buf)));
+ xmmY1 = _mm_loadl_epi64(reinterpret_cast<__m128i*>(reinterpret_cast<uint8*>(kCoefficientsRgbY) + 8 * *y_buf));
+ xmmY1 = _mm_adds_epi16(xmmY1, xmm0);
+ xmmY1 = _mm_srai_epi16(xmmY1, 6);
+ xmmY1 = _mm_packus_epi16(xmmY1, xmmY1);
+ *reinterpret_cast<uint32*>(rgb_buf) = _mm_cvtsi128_si32(xmmY1);
+ }
+}
+
+static void ScaleYUVToRGB32Row_SSE2(const uint8* y_buf,
+ const uint8* u_buf,
+ const uint8* v_buf,
+ uint8* rgb_buf,
+ int width,
+ int source_dx) {
+ __m128i xmm0, xmmY1, xmmY2;
+ __m128 xmmY;
+ uint8 u, v, y;
+ int x = 0;
+
+ while (width >= 2) {
+ u = u_buf[x >> 17];
+ v = v_buf[x >> 17];
+ y = y_buf[x >> 16];
+ x += source_dx;
+
+ xmm0 = _mm_adds_epi16(_mm_loadl_epi64(reinterpret_cast<__m128i*>(kCoefficientsRgbU + 8 * u)),
+ _mm_loadl_epi64(reinterpret_cast<__m128i*>(kCoefficientsRgbV + 8 * v)));
+ xmmY1 = _mm_loadl_epi64(reinterpret_cast<__m128i*>(reinterpret_cast<uint8*>(kCoefficientsRgbY) + 8 * y));
+ xmmY1 = _mm_adds_epi16(xmmY1, xmm0);
+
+ y = y_buf[x >> 16];
+ x += source_dx;
+
+ xmmY2 = _mm_loadl_epi64(reinterpret_cast<__m128i*>(reinterpret_cast<uint8*>(kCoefficientsRgbY) + 8 * y));
+ xmmY2 = _mm_adds_epi16(xmmY2, xmm0);
+
+ xmmY = _mm_shuffle_ps(_mm_castsi128_ps(xmmY1), _mm_castsi128_ps(xmmY2),
+ 0x44);
+ xmmY1 = _mm_srai_epi16(_mm_castps_si128(xmmY), 6);
+ xmmY1 = _mm_packus_epi16(xmmY1, xmmY1);
+
+ _mm_storel_epi64(reinterpret_cast<__m128i*>(rgb_buf), xmmY1);
+ rgb_buf += 8;
+ width -= 2;
+ }
+
+ if (width) {
+ u = u_buf[x >> 17];
+ v = v_buf[x >> 17];
+ y = y_buf[x >> 16];
+
+ xmm0 = _mm_adds_epi16(_mm_loadl_epi64(reinterpret_cast<__m128i*>(kCoefficientsRgbU + 8 * u)),
+ _mm_loadl_epi64(reinterpret_cast<__m128i*>(kCoefficientsRgbV + 8 * v)));
+ xmmY1 = _mm_loadl_epi64(reinterpret_cast<__m128i*>(reinterpret_cast<uint8*>(kCoefficientsRgbY) + 8 * y));
+ xmmY1 = _mm_adds_epi16(xmmY1, xmm0);
+ xmmY1 = _mm_srai_epi16(xmmY1, 6);
+ xmmY1 = _mm_packus_epi16(xmmY1, xmmY1);
+ *reinterpret_cast<uint32*>(rgb_buf) = _mm_cvtsi128_si32(xmmY1);
+ }
+}
+
+static void LinearScaleYUVToRGB32Row_SSE2(const uint8* y_buf,
+ const uint8* u_buf,
+ const uint8* v_buf,
+ uint8* rgb_buf,
+ int width,
+ int source_dx) {
+ __m128i xmm0, xmmY1, xmmY2;
+ __m128 xmmY;
+ uint8 u0, u1, v0, v1, y0, y1;
+ uint32 uv_frac, y_frac, u, v, y;
+ int x = 0;
+
+ if (source_dx >= 0x20000) {
+ x = 32768;
+ }
+
+ while(width >= 2) {
+ u0 = u_buf[x >> 17];
+ u1 = u_buf[(x >> 17) + 1];
+ v0 = v_buf[x >> 17];
+ v1 = v_buf[(x >> 17) + 1];
+ y0 = y_buf[x >> 16];
+ y1 = y_buf[(x >> 16) + 1];
+ uv_frac = (x & 0x1fffe);
+ y_frac = (x & 0xffff);
+ u = (uv_frac * u1 + (uv_frac ^ 0x1fffe) * u0) >> 17;
+ v = (uv_frac * v1 + (uv_frac ^ 0x1fffe) * v0) >> 17;
+ y = (y_frac * y1 + (y_frac ^ 0xffff) * y0) >> 16;
+ x += source_dx;
+
+ xmm0 = _mm_adds_epi16(_mm_loadl_epi64(reinterpret_cast<__m128i*>(kCoefficientsRgbU + 8 * u)),
+ _mm_loadl_epi64(reinterpret_cast<__m128i*>(kCoefficientsRgbV + 8 * v)));
+ xmmY1 = _mm_loadl_epi64(reinterpret_cast<__m128i*>(reinterpret_cast<uint8*>(kCoefficientsRgbY) + 8 * y));
+ xmmY1 = _mm_adds_epi16(xmmY1, xmm0);
+
+ y0 = y_buf[x >> 16];
+ y1 = y_buf[(x >> 16) + 1];
+ y_frac = (x & 0xffff);
+ y = (y_frac * y1 + (y_frac ^ 0xffff) * y0) >> 16;
+ x += source_dx;
+
+ xmmY2 = _mm_loadl_epi64(reinterpret_cast<__m128i*>(reinterpret_cast<uint8*>(kCoefficientsRgbY) + 8 * y));
+ xmmY2 = _mm_adds_epi16(xmmY2, xmm0);
+
+ xmmY = _mm_shuffle_ps(_mm_castsi128_ps(xmmY1), _mm_castsi128_ps(xmmY2),
+ 0x44);
+ xmmY1 = _mm_srai_epi16(_mm_castps_si128(xmmY), 6);
+ xmmY1 = _mm_packus_epi16(xmmY1, xmmY1);
+
+ _mm_storel_epi64(reinterpret_cast<__m128i*>(rgb_buf), xmmY1);
+ rgb_buf += 8;
+ width -= 2;
+ }
+
+ if (width) {
+ u = u_buf[x >> 17];
+ v = v_buf[x >> 17];
+ y = y_buf[x >> 16];
+
+ xmm0 = _mm_adds_epi16(_mm_loadl_epi64(reinterpret_cast<__m128i*>(kCoefficientsRgbU + 8 * u)),
+ _mm_loadl_epi64(reinterpret_cast<__m128i*>(kCoefficientsRgbV + 8 * v)));
+ xmmY1 = _mm_loadl_epi64(reinterpret_cast<__m128i*>(reinterpret_cast<uint8*>(kCoefficientsRgbY) + 8 * y));
+
+ xmmY1 = _mm_adds_epi16(xmmY1, xmm0);
+ xmmY1 = _mm_srai_epi16(xmmY1, 6);
+ xmmY1 = _mm_packus_epi16(xmmY1, xmmY1);
+ *reinterpret_cast<uint32*>(rgb_buf) = _mm_cvtsi128_si32(xmmY1);
+ }
+}
+
+void FastConvertYUVToRGB32Row(const uint8* y_buf,
+ const uint8* u_buf,
+ const uint8* v_buf,
+ uint8* rgb_buf,
+ int width) {
+ FastConvertYUVToRGB32Row_SSE2(y_buf, u_buf, v_buf, rgb_buf, width);
+}
+
+void ScaleYUVToRGB32Row(const uint8* y_buf,
+ const uint8* u_buf,
+ const uint8* v_buf,
+ uint8* rgb_buf,
+ int width,
+ int source_dx) {
+ ScaleYUVToRGB32Row_SSE2(y_buf, u_buf, v_buf, rgb_buf, width, source_dx);
+}
+
+void LinearScaleYUVToRGB32Row(const uint8* y_buf,
+ const uint8* u_buf,
+ const uint8* v_buf,
+ uint8* rgb_buf,
+ int width,
+ int source_dx) {
+ LinearScaleYUVToRGB32Row_SSE2(y_buf, u_buf, v_buf, rgb_buf, width,
+ source_dx);
+}
+
+} // extern "C"