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 mc_rep_func(avg, sz, hsz, hszb, dir, opt, type, f_sz, bpp) \
45 static av_always_inline void \
46 ff_vp9_##avg##_8tap_1d_##dir##_##sz##_##bpp##_##opt(uint8_t *dst, ptrdiff_t dst_stride, \
47 const uint8_t *src, ptrdiff_t src_stride, \
48 int h, const type (*filter)[f_sz]) \
50 ff_vp9_##avg##_8tap_1d_##dir##_##hsz##_##bpp##_##opt(dst, dst_stride, src, \
51 src_stride, h, filter); \
52 ff_vp9_##avg##_8tap_1d_##dir##_##hsz##_##bpp##_##opt(dst + hszb, dst_stride, src + hszb, \
53 src_stride, h, filter); \
56 #define mc_rep_funcs(sz, hsz, hszb, opt, type, fsz, bpp) \
57 mc_rep_func(put, sz, hsz, hszb, h, opt, type, fsz, bpp); \
58 mc_rep_func(avg, sz, hsz, hszb, h, opt, type, fsz, bpp); \
59 mc_rep_func(put, sz, hsz, hszb, v, opt, type, fsz, bpp); \
60 mc_rep_func(avg, sz, hsz, hszb, v, opt, type, fsz, bpp)
62 #define filter_8tap_1d_fn(op, sz, f, f_opt, fname, dir, dvar, bpp, opt) \
63 static void op##_8tap_##fname##_##sz##dir##_##bpp##_##opt(uint8_t *dst, ptrdiff_t dst_stride, \
64 const uint8_t *src, ptrdiff_t src_stride, \
65 int h, int mx, int my) \
67 ff_vp9_##op##_8tap_1d_##dir##_##sz##_##bpp##_##opt(dst, dst_stride, src, src_stride, \
68 h, ff_filters_##f_opt[f][dvar - 1]); \
71 #define filters_8tap_1d_fn(op, sz, dir, dvar, bpp, opt, f_opt) \
72 filter_8tap_1d_fn(op, sz, FILTER_8TAP_REGULAR, f_opt, regular, dir, dvar, bpp, opt) \
73 filter_8tap_1d_fn(op, sz, FILTER_8TAP_SHARP, f_opt, sharp, dir, dvar, bpp, opt) \
74 filter_8tap_1d_fn(op, sz, FILTER_8TAP_SMOOTH, f_opt, smooth, dir, dvar, bpp, opt)
76 #define filters_8tap_1d_fn2(op, sz, bpp, opt, f_opt) \
77 filters_8tap_1d_fn(op, sz, h, mx, bpp, opt, f_opt) \
78 filters_8tap_1d_fn(op, sz, v, my, bpp, opt, f_opt)
80 #define filters_8tap_1d_fn3(op, bpp, opt4, opt8, f_opt) \
81 filters_8tap_1d_fn2(op, 64, bpp, opt8, f_opt) \
82 filters_8tap_1d_fn2(op, 32, bpp, opt8, f_opt) \
83 filters_8tap_1d_fn2(op, 16, bpp, opt8, f_opt) \
84 filters_8tap_1d_fn2(op, 8, bpp, opt8, f_opt) \
85 filters_8tap_1d_fn2(op, 4, bpp, opt4, f_opt)
87 #define filter_8tap_2d_fn(op, sz, f, f_opt, fname, align, bpp, bytes, opt) \
88 static void op##_8tap_##fname##_##sz##hv_##bpp##_##opt(uint8_t *dst, ptrdiff_t dst_stride, \
89 const uint8_t *src, ptrdiff_t src_stride, \
90 int h, int mx, int my) \
92 LOCAL_ALIGNED_##align(uint8_t, temp, [71 * 64 * bytes]); \
93 ff_vp9_put_8tap_1d_h_##sz##_##bpp##_##opt(temp, 64 * bytes, src - 3 * src_stride, \
95 ff_filters_##f_opt[f][mx - 1]); \
96 ff_vp9_##op##_8tap_1d_v_##sz##_##bpp##_##opt(dst, dst_stride, temp + 3 * bytes * 64, \
98 ff_filters_##f_opt[f][my - 1]); \
101 #define filters_8tap_2d_fn(op, sz, align, bpp, bytes, opt, f_opt) \
102 filter_8tap_2d_fn(op, sz, FILTER_8TAP_REGULAR, f_opt, regular, align, bpp, bytes, opt) \
103 filter_8tap_2d_fn(op, sz, FILTER_8TAP_SHARP, f_opt, sharp, align, bpp, bytes, opt) \
104 filter_8tap_2d_fn(op, sz, FILTER_8TAP_SMOOTH, f_opt, smooth, align, bpp, bytes, opt)
106 #define filters_8tap_2d_fn2(op, align, bpp, bytes, opt4, opt8, f_opt) \
107 filters_8tap_2d_fn(op, 64, align, bpp, bytes, opt8, f_opt) \
108 filters_8tap_2d_fn(op, 32, align, bpp, bytes, opt8, f_opt) \
109 filters_8tap_2d_fn(op, 16, align, bpp, bytes, opt8, f_opt) \
110 filters_8tap_2d_fn(op, 8, align, bpp, bytes, opt8, f_opt) \
111 filters_8tap_2d_fn(op, 4, align, bpp, bytes, opt4, f_opt)
113 #define init_fpel_func(idx1, idx2, sz, type, bpp, opt) \
114 dsp->mc[idx1][FILTER_8TAP_SMOOTH ][idx2][0][0] = \
115 dsp->mc[idx1][FILTER_8TAP_REGULAR][idx2][0][0] = \
116 dsp->mc[idx1][FILTER_8TAP_SHARP ][idx2][0][0] = \
117 dsp->mc[idx1][FILTER_BILINEAR ][idx2][0][0] = ff_vp9_##type##sz##bpp##_##opt
119 #define init_subpel1(idx1, idx2, idxh, idxv, sz, dir, type, bpp, opt) \
120 dsp->mc[idx1][FILTER_8TAP_SMOOTH ][idx2][idxh][idxv] = \
121 type##_8tap_smooth_##sz##dir##_##bpp##_##opt; \
122 dsp->mc[idx1][FILTER_8TAP_REGULAR][idx2][idxh][idxv] = \
123 type##_8tap_regular_##sz##dir##_##bpp##_##opt; \
124 dsp->mc[idx1][FILTER_8TAP_SHARP ][idx2][idxh][idxv] = \
125 type##_8tap_sharp_##sz##dir##_##bpp##_##opt
127 #define init_subpel2(idx1, idx2, sz, type, bpp, opt) \
128 init_subpel1(idx1, idx2, 1, 1, sz, hv, type, bpp, opt); \
129 init_subpel1(idx1, idx2, 0, 1, sz, v, type, bpp, opt); \
130 init_subpel1(idx1, idx2, 1, 0, sz, h, type, bpp, opt)
132 #define init_subpel3_32_64(idx, type, bpp, opt) \
133 init_subpel2(0, idx, 64, type, bpp, opt); \
134 init_subpel2(1, idx, 32, type, bpp, opt)
136 #define init_subpel3_8to64(idx, type, bpp, opt) \
137 init_subpel3_32_64(idx, type, bpp, opt); \
138 init_subpel2(2, idx, 16, type, bpp, opt); \
139 init_subpel2(3, idx, 8, type, bpp, opt)
141 #define init_subpel3(idx, type, bpp, opt) \
142 init_subpel3_8to64(idx, type, bpp, opt); \
143 init_subpel2(4, idx, 4, type, bpp, opt)
145 void ff_vp9dsp_init_10bpp_x86(VP9DSPContext *dsp);
146 void ff_vp9dsp_init_12bpp_x86(VP9DSPContext *dsp);
147 void ff_vp9dsp_init_16bpp_x86(VP9DSPContext *dsp);
149 #endif /* AVCODEC_X86_VP9DSP_INIT_H */