Drop unnecessary libavutil/x86/asm.h #includes
[ffmpeg.git] / libavcodec / x86 / vp9dsp_init.c
1 /*
2  * VP9 SIMD optimizations
3  *
4  * Copyright (c) 2013 Ronald S. Bultje <rsbultje@gmail.com>
5  *
6  * This file is part of Libav.
7  *
8  * Libav is free software; you can redistribute it and/or
9  * modify it under the terms of the GNU Lesser General Public
10  * License as published by the Free Software Foundation; either
11  * version 2.1 of the License, or (at your option) any later version.
12  *
13  * Libav is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
16  * Lesser General Public License for more details.
17  *
18  * You should have received a copy of the GNU Lesser General Public
19  * License along with Libav; if not, write to the Free Software
20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21  */
22
23 #include "libavutil/attributes.h"
24 #include "libavutil/cpu.h"
25 #include "libavutil/internal.h"
26 #include "libavutil/mem.h"
27 #include "libavutil/x86/cpu.h"
28 #include "libavcodec/vp9.h"
29
30 #if HAVE_YASM
31
32 #define fpel_func(avg, sz, opt)                                         \
33 void ff_ ## avg ## sz ## _ ## opt(uint8_t *dst, const uint8_t *src,     \
34                                   ptrdiff_t dst_stride,                 \
35                                   ptrdiff_t src_stride,                 \
36                                   int h, int mx, int my)
37
38 fpel_func(put,  4, mmx);
39 fpel_func(put,  8, mmx);
40 fpel_func(put, 16, sse);
41 fpel_func(put, 32, sse);
42 fpel_func(put, 64, sse);
43 fpel_func(avg,  4, sse);
44 fpel_func(avg,  8, sse);
45 fpel_func(avg, 16, sse2);
46 fpel_func(avg, 32, sse2);
47 fpel_func(avg, 64, sse2);
48 #undef fpel_func
49
50 #define mc_func(avg, sz, dir, opt)                                          \
51 void                                                                        \
52 ff_ ## avg ## _8tap_1d_ ## dir ## _ ## sz ## _ ## opt(uint8_t *dst,         \
53                                                       const uint8_t *src,   \
54                                                       ptrdiff_t dst_stride, \
55                                                       ptrdiff_t src_stride, \
56                                                       int h,                \
57                                                       const int8_t (*filter)[16])
58
59 #define mc_funcs(sz)            \
60     mc_func(put, sz, h, ssse3); \
61     mc_func(avg, sz, h, ssse3); \
62     mc_func(put, sz, v, ssse3); \
63     mc_func(avg, sz, v, ssse3)
64
65 mc_funcs(4);
66 mc_funcs(8);
67
68 #undef mc_funcs
69 #undef mc_func
70
71 #define mc_rep_func(avg, sz, hsz, dir, opt)                                 \
72 static av_always_inline void                                                \
73 ff_ ## avg ## _8tap_1d_ ## dir ## _ ## sz ## _ ## opt(uint8_t *dst,         \
74                                                       const uint8_t *src,   \
75                                                       ptrdiff_t dst_stride, \
76                                                       ptrdiff_t src_stride, \
77                                                       int h,                \
78                                                       const int8_t (*filter)[16]) \
79 {                                                                           \
80     ff_ ## avg ## _8tap_1d_ ## dir ## _ ## hsz ## _ ## opt(dst, src,        \
81                                                            dst_stride,      \
82                                                            src_stride,      \
83                                                            h,               \
84                                                            filter);         \
85     ff_ ## avg ## _8tap_1d_ ## dir ## _ ## hsz ## _ ## opt(dst + hsz,       \
86                                                            src + hsz,       \
87                                                            dst_stride,      \
88                                                            src_stride,      \
89                                                            h, filter);      \
90 }
91
92 #define mc_rep_funcs(sz, hsz)            \
93     mc_rep_func(put, sz, hsz, h, ssse3); \
94     mc_rep_func(avg, sz, hsz, h, ssse3); \
95     mc_rep_func(put, sz, hsz, v, ssse3); \
96     mc_rep_func(avg, sz, hsz, v, ssse3)
97
98 mc_rep_funcs(16, 8);
99 mc_rep_funcs(32, 16);
100 mc_rep_funcs(64, 32);
101
102 #undef mc_rep_funcs
103 #undef mc_rep_func
104
105 extern const int8_t ff_filters_ssse3[3][15][4][16];
106
107 #define filter_8tap_2d_fn(op, sz, f, fname)                             \
108 static void                                                             \
109 op ## _8tap_ ## fname ## _ ## sz ## hv_ssse3(uint8_t *dst,              \
110                                              const uint8_t *src,        \
111                                              ptrdiff_t dst_stride,      \
112                                              ptrdiff_t src_stride,      \
113                                              int h, int mx, int my)     \
114 {                                                                       \
115     LOCAL_ALIGNED_16(uint8_t, temp, [71 * 64]);                         \
116     ff_put_8tap_1d_h_ ## sz ## _ssse3(temp, src - 3 * src_stride,       \
117                                       64, src_stride,                   \
118                                       h + 7,                            \
119                                       ff_filters_ssse3[f][mx - 1]);     \
120     ff_ ## op ## _8tap_1d_v_ ## sz ## _ssse3(dst, temp + 3 * 64,        \
121                                              dst_stride, 64,            \
122                                              h,                         \
123                                              ff_filters_ssse3[f][my - 1]); \
124 }
125
126 #define filters_8tap_2d_fn(op, sz)                          \
127     filter_8tap_2d_fn(op, sz, FILTER_8TAP_REGULAR, regular) \
128     filter_8tap_2d_fn(op, sz, FILTER_8TAP_SHARP, sharp)     \
129     filter_8tap_2d_fn(op, sz, FILTER_8TAP_SMOOTH, smooth)
130
131 #define filters_8tap_2d_fn2(op) \
132     filters_8tap_2d_fn(op, 64)  \
133     filters_8tap_2d_fn(op, 32)  \
134     filters_8tap_2d_fn(op, 16)  \
135     filters_8tap_2d_fn(op, 8)   \
136     filters_8tap_2d_fn(op, 4)
137
138 filters_8tap_2d_fn2(put)
139 filters_8tap_2d_fn2(avg)
140
141 #undef filters_8tap_2d_fn2
142 #undef filters_8tap_2d_fn
143 #undef filter_8tap_2d_fn
144
145 #define filter_8tap_1d_fn(op, sz, f, fname, dir, dvar)                  \
146 static void                                                             \
147 op ## _8tap_ ## fname ## _ ## sz ## dir ## _ssse3(uint8_t *dst,         \
148                                                   const uint8_t *src,   \
149                                                   ptrdiff_t dst_stride, \
150                                                   ptrdiff_t src_stride, \
151                                                   int h, int mx,        \
152                                                   int my)               \
153 {                                                                       \
154     ff_ ## op ## _8tap_1d_ ## dir ## _ ## sz ## _ssse3(dst, src,        \
155                                                        dst_stride,      \
156                                                        src_stride, h,   \
157                                                        ff_filters_ssse3[f][dvar - 1]); \
158 }
159
160 #define filters_8tap_1d_fn(op, sz, dir, dvar)                          \
161     filter_8tap_1d_fn(op, sz, FILTER_8TAP_REGULAR, regular, dir, dvar) \
162     filter_8tap_1d_fn(op, sz, FILTER_8TAP_SHARP, sharp, dir, dvar)     \
163     filter_8tap_1d_fn(op, sz, FILTER_8TAP_SMOOTH, smooth, dir, dvar)
164
165 #define filters_8tap_1d_fn2(op, sz)             \
166     filters_8tap_1d_fn(op, sz, h, mx)           \
167     filters_8tap_1d_fn(op, sz, v, my)
168
169 #define filters_8tap_1d_fn3(op) \
170     filters_8tap_1d_fn2(op, 64) \
171     filters_8tap_1d_fn2(op, 32) \
172     filters_8tap_1d_fn2(op, 16) \
173     filters_8tap_1d_fn2(op,  8) \
174     filters_8tap_1d_fn2(op,  4)
175
176 filters_8tap_1d_fn3(put)
177 filters_8tap_1d_fn3(avg)
178
179 #undef filters_8tap_1d_fn
180 #undef filters_8tap_1d_fn2
181 #undef filters_8tap_1d_fn3
182 #undef filter_8tap_1d_fn
183
184 #endif /* HAVE_YASM */
185
186 av_cold void ff_vp9dsp_init_x86(VP9DSPContext *dsp)
187 {
188 #if HAVE_YASM
189     int cpu_flags = av_get_cpu_flags();
190
191 #define init_fpel(idx1, idx2, sz, type, opt)                            \
192     dsp->mc[idx1][FILTER_8TAP_SMOOTH ][idx2][0][0] =                    \
193     dsp->mc[idx1][FILTER_8TAP_REGULAR][idx2][0][0] =                    \
194     dsp->mc[idx1][FILTER_8TAP_SHARP  ][idx2][0][0] =                    \
195     dsp->mc[idx1][FILTER_BILINEAR    ][idx2][0][0] = ff_ ## type ## sz ## _ ## opt
196
197
198 #define init_subpel1(idx1, idx2, idxh, idxv, sz, dir, type, opt) \
199     dsp->mc[idx1][FILTER_8TAP_SMOOTH][idx2][idxh][idxv]  = type ## _8tap_smooth_  ## sz ## dir ## _ ## opt; \
200     dsp->mc[idx1][FILTER_8TAP_REGULAR][idx2][idxh][idxv] = type ## _8tap_regular_ ## sz ## dir ## _ ## opt; \
201     dsp->mc[idx1][FILTER_8TAP_SHARP][idx2][idxh][idxv]   = type ## _8tap_sharp_   ## sz ## dir ## _ ## opt
202
203 #define init_subpel2(idx, idxh, idxv, dir, type, opt)     \
204     init_subpel1(0, idx, idxh, idxv, 64, dir, type, opt); \
205     init_subpel1(1, idx, idxh, idxv, 32, dir, type, opt); \
206     init_subpel1(2, idx, idxh, idxv, 16, dir, type, opt); \
207     init_subpel1(3, idx, idxh, idxv,  8, dir, type, opt); \
208     init_subpel1(4, idx, idxh, idxv,  4, dir, type, opt)
209
210 #define init_subpel3(idx, type, opt)        \
211     init_subpel2(idx, 1, 1, hv, type, opt); \
212     init_subpel2(idx, 0, 1,  v, type, opt); \
213     init_subpel2(idx, 1, 0,  h, type, opt)
214
215     if (EXTERNAL_MMX(cpu_flags)) {
216         init_fpel(4, 0,  4, put, mmx);
217         init_fpel(3, 0,  8, put, mmx);
218     }
219
220     if (EXTERNAL_SSE(cpu_flags)) {
221         init_fpel(2, 0, 16, put, sse);
222         init_fpel(1, 0, 32, put, sse);
223         init_fpel(0, 0, 64, put, sse);
224         init_fpel(4, 1,  4, avg, sse);
225         init_fpel(3, 1,  8, avg, sse);
226     }
227
228     if (EXTERNAL_SSE2(cpu_flags)) {
229         init_fpel(2, 1, 16, avg, sse2);
230         init_fpel(1, 1, 32, avg, sse2);
231         init_fpel(0, 1, 64, avg, sse2);
232     }
233
234     if (EXTERNAL_SSSE3(cpu_flags)) {
235         init_subpel3(0, put, ssse3);
236         init_subpel3(1, avg, ssse3);
237     }
238
239 #undef init_fpel
240 #undef init_subpel1
241 #undef init_subpel2
242 #undef init_subpel3
243
244 #endif /* HAVE_YASM */
245 }