2 * VP9 SIMD optimizations
4 * Copyright (c) 2013 Ronald S. Bultje <rsbultje gmail com>
6 * This file is part of FFmpeg.
8 * FFmpeg 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.
13 * FFmpeg 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.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 #ifndef AVCODEC_X86_VP9DSP_INIT_H
24 #define AVCODEC_X86_VP9DSP_INIT_H
26 #include "libavcodec/vp9dsp.h"
28 #define decl_fpel_func(avg, sz, bpp, opt) \
29 void ff_vp9_##avg##sz##bpp##_##opt(uint8_t *dst, ptrdiff_t dst_stride, \
30 const uint8_t *src, ptrdiff_t src_stride, \
31 int h, int mx, int my)
33 #define decl_mc_func(avg, sz, dir, opt, type, f_sz, bpp) \
34 void ff_vp9_##avg##_8tap_1d_##dir##_##sz##_##bpp##_##opt(uint8_t *dst, ptrdiff_t dst_stride, \
35 const uint8_t *src, ptrdiff_t src_stride, \
36 int h, const type (*filter)[f_sz])
38 #define decl_mc_funcs(sz, opt, type, fsz, bpp) \
39 decl_mc_func(put, sz, h, opt, type, fsz, bpp); \
40 decl_mc_func(avg, sz, h, opt, type, fsz, bpp); \
41 decl_mc_func(put, sz, v, opt, type, fsz, bpp); \
42 decl_mc_func(avg, sz, v, opt, type, fsz, bpp)
44 #define decl_ipred_fn(type, sz, bpp, opt) \
45 void ff_vp9_ipred_##type##_##sz##x##sz##_##bpp##_##opt(uint8_t *dst, \
50 #define decl_ipred_fns(type, bpp, opt4, opt8_16_32) \
51 decl_ipred_fn(type, 4, bpp, opt4); \
52 decl_ipred_fn(type, 8, bpp, opt8_16_32); \
53 decl_ipred_fn(type, 16, bpp, opt8_16_32); \
54 decl_ipred_fn(type, 32, bpp, opt8_16_32)
56 #define mc_rep_func(avg, sz, hsz, hszb, dir, opt, type, f_sz, bpp) \
57 static av_always_inline void \
58 ff_vp9_##avg##_8tap_1d_##dir##_##sz##_##bpp##_##opt(uint8_t *dst, ptrdiff_t dst_stride, \
59 const uint8_t *src, ptrdiff_t src_stride, \
60 int h, const type (*filter)[f_sz]) \
62 ff_vp9_##avg##_8tap_1d_##dir##_##hsz##_##bpp##_##opt(dst, dst_stride, src, \
63 src_stride, h, filter); \
64 ff_vp9_##avg##_8tap_1d_##dir##_##hsz##_##bpp##_##opt(dst + hszb, dst_stride, src + hszb, \
65 src_stride, h, filter); \
68 #define mc_rep_funcs(sz, hsz, hszb, opt, type, fsz, bpp) \
69 mc_rep_func(put, sz, hsz, hszb, h, opt, type, fsz, bpp); \
70 mc_rep_func(avg, sz, hsz, hszb, h, opt, type, fsz, bpp); \
71 mc_rep_func(put, sz, hsz, hszb, v, opt, type, fsz, bpp); \
72 mc_rep_func(avg, sz, hsz, hszb, v, opt, type, fsz, bpp)
74 #define filter_8tap_1d_fn(op, sz, f, f_opt, fname, dir, dvar, bpp, opt) \
75 static void op##_8tap_##fname##_##sz##dir##_##bpp##_##opt(uint8_t *dst, ptrdiff_t dst_stride, \
76 const uint8_t *src, ptrdiff_t src_stride, \
77 int h, int mx, int my) \
79 ff_vp9_##op##_8tap_1d_##dir##_##sz##_##bpp##_##opt(dst, dst_stride, src, src_stride, \
80 h, ff_filters_##f_opt[f][dvar - 1]); \
83 #define filters_8tap_1d_fn(op, sz, dir, dvar, bpp, opt, f_opt) \
84 filter_8tap_1d_fn(op, sz, FILTER_8TAP_REGULAR, f_opt, regular, dir, dvar, bpp, opt) \
85 filter_8tap_1d_fn(op, sz, FILTER_8TAP_SHARP, f_opt, sharp, dir, dvar, bpp, opt) \
86 filter_8tap_1d_fn(op, sz, FILTER_8TAP_SMOOTH, f_opt, smooth, dir, dvar, bpp, opt)
88 #define filters_8tap_1d_fn2(op, sz, bpp, opt, f_opt) \
89 filters_8tap_1d_fn(op, sz, h, mx, bpp, opt, f_opt) \
90 filters_8tap_1d_fn(op, sz, v, my, bpp, opt, f_opt)
92 #define filters_8tap_1d_fn3(op, bpp, opt4, opt8, f_opt) \
93 filters_8tap_1d_fn2(op, 64, bpp, opt8, f_opt) \
94 filters_8tap_1d_fn2(op, 32, bpp, opt8, f_opt) \
95 filters_8tap_1d_fn2(op, 16, bpp, opt8, f_opt) \
96 filters_8tap_1d_fn2(op, 8, bpp, opt8, f_opt) \
97 filters_8tap_1d_fn2(op, 4, bpp, opt4, f_opt)
99 #define filter_8tap_2d_fn(op, sz, f, f_opt, fname, align, bpp, bytes, opt) \
100 static void op##_8tap_##fname##_##sz##hv_##bpp##_##opt(uint8_t *dst, ptrdiff_t dst_stride, \
101 const uint8_t *src, ptrdiff_t src_stride, \
102 int h, int mx, int my) \
104 LOCAL_ALIGNED_##align(uint8_t, temp, [71 * 64 * bytes]); \
105 ff_vp9_put_8tap_1d_h_##sz##_##bpp##_##opt(temp, 64 * bytes, src - 3 * src_stride, \
107 ff_filters_##f_opt[f][mx - 1]); \
108 ff_vp9_##op##_8tap_1d_v_##sz##_##bpp##_##opt(dst, dst_stride, temp + 3 * bytes * 64, \
110 ff_filters_##f_opt[f][my - 1]); \
113 #define filters_8tap_2d_fn(op, sz, align, bpp, bytes, opt, f_opt) \
114 filter_8tap_2d_fn(op, sz, FILTER_8TAP_REGULAR, f_opt, regular, align, bpp, bytes, opt) \
115 filter_8tap_2d_fn(op, sz, FILTER_8TAP_SHARP, f_opt, sharp, align, bpp, bytes, opt) \
116 filter_8tap_2d_fn(op, sz, FILTER_8TAP_SMOOTH, f_opt, smooth, align, bpp, bytes, opt)
118 #define filters_8tap_2d_fn2(op, align, bpp, bytes, opt4, opt8, f_opt) \
119 filters_8tap_2d_fn(op, 64, align, bpp, bytes, opt8, f_opt) \
120 filters_8tap_2d_fn(op, 32, align, bpp, bytes, opt8, f_opt) \
121 filters_8tap_2d_fn(op, 16, align, bpp, bytes, opt8, f_opt) \
122 filters_8tap_2d_fn(op, 8, align, bpp, bytes, opt8, f_opt) \
123 filters_8tap_2d_fn(op, 4, align, bpp, bytes, opt4, f_opt)
125 #define init_fpel_func(idx1, idx2, sz, type, bpp, opt) \
126 dsp->mc[idx1][FILTER_8TAP_SMOOTH ][idx2][0][0] = \
127 dsp->mc[idx1][FILTER_8TAP_REGULAR][idx2][0][0] = \
128 dsp->mc[idx1][FILTER_8TAP_SHARP ][idx2][0][0] = \
129 dsp->mc[idx1][FILTER_BILINEAR ][idx2][0][0] = ff_vp9_##type##sz##bpp##_##opt
131 #define init_subpel1(idx1, idx2, idxh, idxv, sz, dir, type, bpp, opt) \
132 dsp->mc[idx1][FILTER_8TAP_SMOOTH ][idx2][idxh][idxv] = \
133 type##_8tap_smooth_##sz##dir##_##bpp##_##opt; \
134 dsp->mc[idx1][FILTER_8TAP_REGULAR][idx2][idxh][idxv] = \
135 type##_8tap_regular_##sz##dir##_##bpp##_##opt; \
136 dsp->mc[idx1][FILTER_8TAP_SHARP ][idx2][idxh][idxv] = \
137 type##_8tap_sharp_##sz##dir##_##bpp##_##opt
139 #define init_subpel2(idx1, idx2, sz, type, bpp, opt) \
140 init_subpel1(idx1, idx2, 1, 1, sz, hv, type, bpp, opt); \
141 init_subpel1(idx1, idx2, 0, 1, sz, v, type, bpp, opt); \
142 init_subpel1(idx1, idx2, 1, 0, sz, h, type, bpp, opt)
144 #define init_subpel3_32_64(idx, type, bpp, opt) \
145 init_subpel2(0, idx, 64, type, bpp, opt); \
146 init_subpel2(1, idx, 32, type, bpp, opt)
148 #define init_subpel3_8to64(idx, type, bpp, opt) \
149 init_subpel3_32_64(idx, type, bpp, opt); \
150 init_subpel2(2, idx, 16, type, bpp, opt); \
151 init_subpel2(3, idx, 8, type, bpp, opt)
153 #define init_subpel3(idx, type, bpp, opt) \
154 init_subpel3_8to64(idx, type, bpp, opt); \
155 init_subpel2(4, idx, 4, type, bpp, opt)
157 #define cat(a, bpp, b) a##bpp##b
159 #define init_ipred_func(type, enum, sz, bpp, opt) \
160 dsp->intra_pred[TX_##sz##X##sz][enum##_PRED] = \
161 cat(ff_vp9_ipred_##type##_##sz##x##sz##_, bpp, _##opt)
163 #define init_8_16_32_ipred_funcs(type, enum, bpp, opt) \
164 init_ipred_func(type, enum, 8, bpp, opt); \
165 init_ipred_func(type, enum, 16, bpp, opt); \
166 init_ipred_func(type, enum, 32, bpp, opt)
168 #define init_ipred_funcs(type, enum, bpp, opt) \
169 init_ipred_func(type, enum, 4, bpp, opt); \
170 init_8_16_32_ipred_funcs(type, enum, bpp, opt)
172 void ff_vp9dsp_init_10bpp_x86(VP9DSPContext *dsp);
173 void ff_vp9dsp_init_12bpp_x86(VP9DSPContext *dsp);
174 void ff_vp9dsp_init_16bpp_x86(VP9DSPContext *dsp);
176 #endif /* AVCODEC_X86_VP9DSP_INIT_H */