1 /*
2 * software RGB to RGB converter
3 * pluralize by software PAL8 to RGB converter
4 * software YUV to YUV converter
5 * software YUV to RGB converter
6 * Written by Nick Kurshev.
7 * palette & YUV & runtime CPU stuff by Michael (michaelni@gmx.at)
8 *
9 * This file is part of FFmpeg.
10 *
11 * FFmpeg is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU Lesser General Public
13 * License as published by the Free Software Foundation; either
14 * version 2.1 of the License, or (at your option) any later version.
15 *
16 * FFmpeg is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 * Lesser General Public License for more details.
20 *
21 * You should have received a copy of the GNU Lesser General Public
22 * License along with FFmpeg; if not, write to the Free Software
23 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 */
25
26 #include <stdint.h>
27
28 #include "config.h"
29 #include "libavutil/attributes.h"
30 #include "libavutil/x86/cpu.h"
31 #include "libavutil/cpu.h"
32 #include "libavutil/bswap.h"
33 #include "libavutil/mem_internal.h"
34
35 #include "libswscale/rgb2rgb.h"
36 #include "libswscale/swscale.h"
37 #include "libswscale/swscale_internal.h"
38
39 #if HAVE_INLINE_ASM
40
41 DECLARE_ASM_CONST(8, uint64_t, mmx_ff) = 0x00000000000000FFULL;
42 DECLARE_ASM_CONST(8, uint64_t, mmx_null) = 0x0000000000000000ULL;
43 DECLARE_ASM_CONST(8, uint64_t, mask32a) = 0xFF000000FF000000ULL;
44 DECLARE_ASM_CONST(8, uint64_t, mask3216br) = 0x00F800F800F800F8ULL;
45 DECLARE_ASM_CONST(8, uint64_t, mask3216g) = 0x0000FC000000FC00ULL;
46 DECLARE_ASM_CONST(8, uint64_t, mask3215g) = 0x0000F8000000F800ULL;
47 DECLARE_ASM_CONST(8, uint64_t, mul3216) = 0x2000000420000004ULL;
48 DECLARE_ASM_CONST(8, uint64_t, mul3215) = 0x2000000820000008ULL;
49 DECLARE_ASM_CONST(8, uint64_t, mask24b) = 0x00FF0000FF0000FFULL;
50 DECLARE_ASM_CONST(8, uint64_t, mask24g) = 0xFF0000FF0000FF00ULL;
51 DECLARE_ASM_CONST(8, uint64_t, mask24r) = 0x0000FF0000FF0000ULL;
52 DECLARE_ASM_CONST(8, uint64_t, mask24l) = 0x0000000000FFFFFFULL;
53 DECLARE_ASM_CONST(8, uint64_t, mask24h) = 0x0000FFFFFF000000ULL;
54 DECLARE_ASM_CONST(8, uint64_t, mask15b) = 0x001F001F001F001FULL; /* 00000000 00011111 xxB */
55 DECLARE_ASM_CONST(8, uint64_t, mask15rg) = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000 RGx */
56 DECLARE_ASM_CONST(8, uint64_t, mask15s) = 0xFFE0FFE0FFE0FFE0ULL;
57 DECLARE_ASM_CONST(8, uint64_t, mask15g) = 0x03E003E003E003E0ULL;
58 DECLARE_ASM_CONST(8, uint64_t, mask15r) = 0x7C007C007C007C00ULL;
59 #define mask16b mask15b
60 DECLARE_ASM_CONST(8, uint64_t, mask16g) = 0x07E007E007E007E0ULL;
61 DECLARE_ASM_CONST(8, uint64_t, mask16r) = 0xF800F800F800F800ULL;
62 DECLARE_ASM_CONST(8, uint64_t, red_16mask) = 0x0000f8000000f800ULL;
63 DECLARE_ASM_CONST(8, uint64_t, green_16mask) = 0x000007e0000007e0ULL;
64 DECLARE_ASM_CONST(8, uint64_t, blue_16mask) = 0x0000001f0000001fULL;
65 DECLARE_ASM_CONST(8, uint64_t, red_15mask) = 0x00007c0000007c00ULL;
66 DECLARE_ASM_CONST(8, uint64_t, green_15mask) = 0x000003e0000003e0ULL;
67 DECLARE_ASM_CONST(8, uint64_t, blue_15mask) = 0x0000001f0000001fULL;
68 DECLARE_ASM_CONST(8, uint64_t, mul15_mid) = 0x4200420042004200ULL;
69 DECLARE_ASM_CONST(8, uint64_t, mul15_hi) = 0x0210021002100210ULL;
70 DECLARE_ASM_CONST(8, uint64_t, mul16_mid) = 0x2080208020802080ULL;
71
72 DECLARE_ALIGNED(8, extern const uint64_t, ff_bgr2YOffset);
73 DECLARE_ALIGNED(8, extern const uint64_t, ff_w1111);
74 DECLARE_ALIGNED(8, extern const uint64_t, ff_bgr2UVOffset);
75
76 #define BY ((int)( 0.098*(1<<RGB2YUV_SHIFT)+0.5))
77 #define BV ((int)(-0.071*(1<<RGB2YUV_SHIFT)+0.5))
78 #define BU ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
79 #define GY ((int)( 0.504*(1<<RGB2YUV_SHIFT)+0.5))
80 #define GV ((int)(-0.368*(1<<RGB2YUV_SHIFT)+0.5))
81 #define GU ((int)(-0.291*(1<<RGB2YUV_SHIFT)+0.5))
82 #define RY ((int)( 0.257*(1<<RGB2YUV_SHIFT)+0.5))
83 #define RV ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
84 #define RU ((int)(-0.148*(1<<RGB2YUV_SHIFT)+0.5))
85
86 // Note: We have C, MMX, MMXEXT, 3DNOW versions, there is no 3DNOW + MMXEXT one.
87
88 #define COMPILE_TEMPLATE_SSE2 0
89 #define COMPILE_TEMPLATE_AVX 0
90
91 // MMXEXT versions
92 #undef RENAME
93 #define RENAME(a) a ## _mmxext
94 #include "rgb2rgb_template.c"
95
96 //SSE2 versions
97 #undef RENAME
98 #undef COMPILE_TEMPLATE_SSE2
99 #define COMPILE_TEMPLATE_SSE2 1
100 #define RENAME(a) a ## _sse2
101 #include "rgb2rgb_template.c"
102
103 //AVX versions
104 #undef RENAME
105 #undef COMPILE_TEMPLATE_AVX
106 #define COMPILE_TEMPLATE_AVX 1
107 #define RENAME(a) a ## _avx
108 #include "rgb2rgb_template.c"
109
110 /*
111 RGB15->RGB16 original by Strepto/Astral
112 ported to gcc & bugfixed : A'rpi
113 MMXEXT, 3DNOW optimization by Nick Kurshev
114 32-bit C version, and and&add trick by Michael Niedermayer
115 */
116
117 #endif /* HAVE_INLINE_ASM */
118
119 void ff_shuffle_bytes_2103_mmxext(const uint8_t *src, uint8_t *dst, int src_size);
120 void ff_shuffle_bytes_2103_ssse3(const uint8_t *src, uint8_t *dst, int src_size);
121 void ff_shuffle_bytes_0321_ssse3(const uint8_t *src, uint8_t *dst, int src_size);
122 void ff_shuffle_bytes_1230_ssse3(const uint8_t *src, uint8_t *dst, int src_size);
123 void ff_shuffle_bytes_3012_ssse3(const uint8_t *src, uint8_t *dst, int src_size);
124 void ff_shuffle_bytes_3210_ssse3(const uint8_t *src, uint8_t *dst, int src_size);
125
126 #if ARCH_X86_64
127 void ff_shuffle_bytes_2103_avx2(const uint8_t *src, uint8_t *dst, int src_size);
128 void ff_shuffle_bytes_0321_avx2(const uint8_t *src, uint8_t *dst, int src_size);
129 void ff_shuffle_bytes_1230_avx2(const uint8_t *src, uint8_t *dst, int src_size);
130 void ff_shuffle_bytes_3012_avx2(const uint8_t *src, uint8_t *dst, int src_size);
131 void ff_shuffle_bytes_3210_avx2(const uint8_t *src, uint8_t *dst, int src_size);
132
133 void ff_uyvytoyuv422_sse2(uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
134 const uint8_t *src, int width, int height,
135 int lumStride, int chromStride, int srcStride);
136 void ff_uyvytoyuv422_avx(uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
137 const uint8_t *src, int width, int height,
138 int lumStride, int chromStride, int srcStride);
139 #endif
140
rgb2rgb_init_x86(void)141 av_cold void rgb2rgb_init_x86(void)
142 {
143 int cpu_flags = av_get_cpu_flags();
144
145 #if HAVE_INLINE_ASM
146 if (INLINE_MMXEXT(cpu_flags))
147 rgb2rgb_init_mmxext();
148 if (INLINE_SSE2(cpu_flags))
149 rgb2rgb_init_sse2();
150 if (INLINE_AVX(cpu_flags))
151 rgb2rgb_init_avx();
152 #endif /* HAVE_INLINE_ASM */
153
154 if (EXTERNAL_MMXEXT(cpu_flags)) {
155 shuffle_bytes_2103 = ff_shuffle_bytes_2103_mmxext;
156 }
157 if (EXTERNAL_SSE2(cpu_flags)) {
158 #if ARCH_X86_64
159 uyvytoyuv422 = ff_uyvytoyuv422_sse2;
160 #endif
161 }
162 if (EXTERNAL_SSSE3(cpu_flags)) {
163 shuffle_bytes_0321 = ff_shuffle_bytes_0321_ssse3;
164 shuffle_bytes_2103 = ff_shuffle_bytes_2103_ssse3;
165 shuffle_bytes_1230 = ff_shuffle_bytes_1230_ssse3;
166 shuffle_bytes_3012 = ff_shuffle_bytes_3012_ssse3;
167 shuffle_bytes_3210 = ff_shuffle_bytes_3210_ssse3;
168 }
169 #if ARCH_X86_64
170 if (EXTERNAL_AVX2_FAST(cpu_flags)) {
171 shuffle_bytes_0321 = ff_shuffle_bytes_0321_avx2;
172 shuffle_bytes_2103 = ff_shuffle_bytes_2103_avx2;
173 shuffle_bytes_1230 = ff_shuffle_bytes_1230_avx2;
174 shuffle_bytes_3012 = ff_shuffle_bytes_3012_avx2;
175 shuffle_bytes_3210 = ff_shuffle_bytes_3210_avx2;
176 }
177 if (EXTERNAL_AVX(cpu_flags)) {
178 uyvytoyuv422 = ff_uyvytoyuv422_avx;
179 }
180 #endif
181 }
182