1 /*
2 * This file is part of FFmpeg.
3 *
4 * FFmpeg is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation; either
7 * version 2.1 of the License, or (at your option) any later version.
8 *
9 * FFmpeg is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Lesser General Public License for more details.
13 *
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with FFmpeg; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18
19 #include "config.h"
20 #include "libswscale/swscale.h"
21 #include "libswscale/swscale_internal.h"
22 #include "libavutil/aarch64/cpu.h"
23
24 #define YUV_TO_RGB_TABLE \
25 c->yuv2rgb_v2r_coeff, \
26 c->yuv2rgb_u2g_coeff, \
27 c->yuv2rgb_v2g_coeff, \
28 c->yuv2rgb_u2b_coeff, \
29
30 #define DECLARE_FF_YUVX_TO_RGBX_FUNCS(ifmt, ofmt) \
31 int ff_##ifmt##_to_##ofmt##_neon(int w, int h, \
32 uint8_t *dst, int linesize, \
33 const uint8_t *srcY, int linesizeY, \
34 const uint8_t *srcU, int linesizeU, \
35 const uint8_t *srcV, int linesizeV, \
36 const int16_t *table, \
37 int y_offset, \
38 int y_coeff); \
39 \
40 static int ifmt##_to_##ofmt##_neon_wrapper(SwsContext *c, const uint8_t *src[], \
41 int srcStride[], int srcSliceY, int srcSliceH, \
42 uint8_t *dst[], int dstStride[]) { \
43 const int16_t yuv2rgb_table[] = { YUV_TO_RGB_TABLE }; \
44 \
45 return ff_##ifmt##_to_##ofmt##_neon(c->srcW, srcSliceH, \
46 dst[0] + srcSliceY * dstStride[0], dstStride[0], \
47 src[0], srcStride[0], \
48 src[1], srcStride[1], \
49 src[2], srcStride[2], \
50 yuv2rgb_table, \
51 c->yuv2rgb_y_offset >> 6, \
52 c->yuv2rgb_y_coeff); \
53 } \
54
55 #define DECLARE_FF_YUVX_TO_ALL_RGBX_FUNCS(yuvx) \
56 DECLARE_FF_YUVX_TO_RGBX_FUNCS(yuvx, argb) \
57 DECLARE_FF_YUVX_TO_RGBX_FUNCS(yuvx, rgba) \
58 DECLARE_FF_YUVX_TO_RGBX_FUNCS(yuvx, abgr) \
59 DECLARE_FF_YUVX_TO_RGBX_FUNCS(yuvx, bgra) \
60
61 DECLARE_FF_YUVX_TO_ALL_RGBX_FUNCS(yuv420p)
DECLARE_FF_YUVX_TO_ALL_RGBX_FUNCS(yuv422p)62 DECLARE_FF_YUVX_TO_ALL_RGBX_FUNCS(yuv422p)
63
64 #define DECLARE_FF_NVX_TO_RGBX_FUNCS(ifmt, ofmt) \
65 int ff_##ifmt##_to_##ofmt##_neon(int w, int h, \
66 uint8_t *dst, int linesize, \
67 const uint8_t *srcY, int linesizeY, \
68 const uint8_t *srcC, int linesizeC, \
69 const int16_t *table, \
70 int y_offset, \
71 int y_coeff); \
72 \
73 static int ifmt##_to_##ofmt##_neon_wrapper(SwsContext *c, const uint8_t *src[], \
74 int srcStride[], int srcSliceY, int srcSliceH, \
75 uint8_t *dst[], int dstStride[]) { \
76 const int16_t yuv2rgb_table[] = { YUV_TO_RGB_TABLE }; \
77 \
78 return ff_##ifmt##_to_##ofmt##_neon(c->srcW, srcSliceH, \
79 dst[0] + srcSliceY * dstStride[0], dstStride[0], \
80 src[0], srcStride[0], src[1], srcStride[1], \
81 yuv2rgb_table, \
82 c->yuv2rgb_y_offset >> 6, \
83 c->yuv2rgb_y_coeff); \
84 } \
85
86 #define DECLARE_FF_NVX_TO_ALL_RGBX_FUNCS(nvx) \
87 DECLARE_FF_NVX_TO_RGBX_FUNCS(nvx, argb) \
88 DECLARE_FF_NVX_TO_RGBX_FUNCS(nvx, rgba) \
89 DECLARE_FF_NVX_TO_RGBX_FUNCS(nvx, abgr) \
90 DECLARE_FF_NVX_TO_RGBX_FUNCS(nvx, bgra) \
91
92 DECLARE_FF_NVX_TO_ALL_RGBX_FUNCS(nv12)
93 DECLARE_FF_NVX_TO_ALL_RGBX_FUNCS(nv21)
94
95 /* We need a 16 pixel width alignment. This constraint can easily be removed
96 * for input reading but for the output which is 4-bytes per pixel (RGBA) the
97 * assembly might be writing as much as 4*15=60 extra bytes at the end of the
98 * line, which won't fit the 32-bytes buffer alignment. */
99 #define SET_FF_NVX_TO_RGBX_FUNC(ifmt, IFMT, ofmt, OFMT, accurate_rnd) do { \
100 if (c->srcFormat == AV_PIX_FMT_##IFMT \
101 && c->dstFormat == AV_PIX_FMT_##OFMT \
102 && !(c->srcH & 1) \
103 && !(c->srcW & 15) \
104 && !accurate_rnd) \
105 c->convert_unscaled = ifmt##_to_##ofmt##_neon_wrapper; \
106 } while (0)
107
108 #define SET_FF_NVX_TO_ALL_RGBX_FUNC(nvx, NVX, accurate_rnd) do { \
109 SET_FF_NVX_TO_RGBX_FUNC(nvx, NVX, argb, ARGB, accurate_rnd); \
110 SET_FF_NVX_TO_RGBX_FUNC(nvx, NVX, rgba, RGBA, accurate_rnd); \
111 SET_FF_NVX_TO_RGBX_FUNC(nvx, NVX, abgr, ABGR, accurate_rnd); \
112 SET_FF_NVX_TO_RGBX_FUNC(nvx, NVX, bgra, BGRA, accurate_rnd); \
113 } while (0)
114
115 static void get_unscaled_swscale_neon(SwsContext *c) {
116 int accurate_rnd = c->flags & SWS_ACCURATE_RND;
117
118 SET_FF_NVX_TO_ALL_RGBX_FUNC(nv12, NV12, accurate_rnd);
119 SET_FF_NVX_TO_ALL_RGBX_FUNC(nv21, NV21, accurate_rnd);
120 SET_FF_NVX_TO_ALL_RGBX_FUNC(yuv420p, YUV420P, accurate_rnd);
121 SET_FF_NVX_TO_ALL_RGBX_FUNC(yuv422p, YUV422P, accurate_rnd);
122 }
123
ff_get_unscaled_swscale_aarch64(SwsContext * c)124 void ff_get_unscaled_swscale_aarch64(SwsContext *c)
125 {
126 int cpu_flags = av_get_cpu_flags();
127 if (have_neon(cpu_flags))
128 get_unscaled_swscale_neon(c);
129 }
130