2 * RV30/40 decoder common dsp functions
3 * Copyright (c) 2007 Mike Melanson, Konstantin Shishkov
4 * Copyright (c) 2011 Janne Grunau
6 * This file is part of Libav.
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.
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.
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
25 * RV30/40 decoder common dsp functions
31 * @name RV30/40 inverse transform functions
35 static av_always_inline void rv34_row_transform(int temp[16], DCTELEM *block)
39 for(i = 0; i < 4; i++){
40 const int z0 = 13*(block[i+8*0] + block[i+8*2]);
41 const int z1 = 13*(block[i+8*0] - block[i+8*2]);
42 const int z2 = 7* block[i+8*1] - 17*block[i+8*3];
43 const int z3 = 17* block[i+8*1] + 7*block[i+8*3];
45 temp[4*i+0] = z0 + z3;
46 temp[4*i+1] = z1 + z2;
47 temp[4*i+2] = z1 - z2;
48 temp[4*i+3] = z0 - z3;
53 * Real Video 3.0/4.0 inverse transform
54 * Code is almost the same as in SVQ3, only scaling is different.
56 static void rv34_inv_transform_c(DCTELEM *block){
60 rv34_row_transform(temp, block);
62 for(i = 0; i < 4; i++){
63 const int z0 = 13*(temp[4*0+i] + temp[4*2+i]) + 0x200;
64 const int z1 = 13*(temp[4*0+i] - temp[4*2+i]) + 0x200;
65 const int z2 = 7* temp[4*1+i] - 17*temp[4*3+i];
66 const int z3 = 17* temp[4*1+i] + 7*temp[4*3+i];
68 block[i*8+0] = (z0 + z3) >> 10;
69 block[i*8+1] = (z1 + z2) >> 10;
70 block[i*8+2] = (z1 - z2) >> 10;
71 block[i*8+3] = (z0 - z3) >> 10;
76 * RealVideo 3.0/4.0 inverse transform for DC block
78 * Code is almost the same as rv34_inv_transform()
79 * but final coefficients are multiplied by 1.5 and have no rounding.
81 static void rv34_inv_transform_noround_c(DCTELEM *block){
85 rv34_row_transform(temp, block);
87 for(i = 0; i < 4; i++){
88 const int z0 = 13*(temp[4*0+i] + temp[4*2+i]);
89 const int z1 = 13*(temp[4*0+i] - temp[4*2+i]);
90 const int z2 = 7* temp[4*1+i] - 17*temp[4*3+i];
91 const int z3 = 17* temp[4*1+i] + 7*temp[4*3+i];
93 block[i*8+0] = ((z0 + z3) * 3) >> 11;
94 block[i*8+1] = ((z1 + z2) * 3) >> 11;
95 block[i*8+2] = ((z1 - z2) * 3) >> 11;
96 block[i*8+3] = ((z0 - z3) * 3) >> 11;
100 /** @} */ // transform
103 av_cold void ff_rv34dsp_init(RV34DSPContext *c, DSPContext* dsp) {
104 c->rv34_inv_transform_tab[0] = rv34_inv_transform_c;
105 c->rv34_inv_transform_tab[1] = rv34_inv_transform_noround_c;
108 ff_rv34dsp_init_neon(c, dsp);