blob: 1d05b2fe08cb869353e12778f2312140a38de2e6
| 1 | /* |
| 2 | * Simple IDCT |
| 3 | * |
| 4 | * Copyright (c) 2001 Michael Niedermayer <michaelni@gmx.at> |
| 5 | * |
| 6 | * This file is part of FFmpeg. |
| 7 | * |
| 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. |
| 12 | * |
| 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. |
| 17 | * |
| 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 |
| 21 | */ |
| 22 | |
| 23 | /** |
| 24 | * @file |
| 25 | * simpleidct in C. |
| 26 | */ |
| 27 | |
| 28 | #include "libavutil/intreadwrite.h" |
| 29 | #include "avcodec.h" |
| 30 | #include "mathops.h" |
| 31 | #include "simple_idct.h" |
| 32 | |
| 33 | #define BIT_DEPTH 8 |
| 34 | #include "simple_idct_template.c" |
| 35 | #undef BIT_DEPTH |
| 36 | |
| 37 | #define BIT_DEPTH 10 |
| 38 | #include "simple_idct_template.c" |
| 39 | |
| 40 | #define EXTRA_SHIFT 2 |
| 41 | #include "simple_idct_template.c" |
| 42 | |
| 43 | #undef EXTRA_SHIFT |
| 44 | #undef BIT_DEPTH |
| 45 | |
| 46 | #define BIT_DEPTH 12 |
| 47 | #include "simple_idct_template.c" |
| 48 | #undef BIT_DEPTH |
| 49 | |
| 50 | /* 2x4x8 idct */ |
| 51 | |
| 52 | #define CN_SHIFT 12 |
| 53 | #define C_FIX(x) ((int)((x) * (1 << CN_SHIFT) + 0.5)) |
| 54 | #define C1 C_FIX(0.6532814824) |
| 55 | #define C2 C_FIX(0.2705980501) |
| 56 | |
| 57 | /* row idct is multiple by 16 * sqrt(2.0), col idct4 is normalized, |
| 58 | and the butterfly must be multiplied by 0.5 * sqrt(2.0) */ |
| 59 | #define C_SHIFT (4+1+12) |
| 60 | |
| 61 | static inline void idct4col_put(uint8_t *dest, ptrdiff_t line_size, const int16_t *col) |
| 62 | { |
| 63 | int c0, c1, c2, c3, a0, a1, a2, a3; |
| 64 | |
| 65 | a0 = col[8*0]; |
| 66 | a1 = col[8*2]; |
| 67 | a2 = col[8*4]; |
| 68 | a3 = col[8*6]; |
| 69 | c0 = ((a0 + a2) * (1 << CN_SHIFT - 1)) + (1 << (C_SHIFT - 1)); |
| 70 | c2 = ((a0 - a2) * (1 << CN_SHIFT - 1)) + (1 << (C_SHIFT - 1)); |
| 71 | c1 = a1 * C1 + a3 * C2; |
| 72 | c3 = a1 * C2 - a3 * C1; |
| 73 | dest[0] = av_clip_uint8((c0 + c1) >> C_SHIFT); |
| 74 | dest += line_size; |
| 75 | dest[0] = av_clip_uint8((c2 + c3) >> C_SHIFT); |
| 76 | dest += line_size; |
| 77 | dest[0] = av_clip_uint8((c2 - c3) >> C_SHIFT); |
| 78 | dest += line_size; |
| 79 | dest[0] = av_clip_uint8((c0 - c1) >> C_SHIFT); |
| 80 | } |
| 81 | |
| 82 | #define BF(k) \ |
| 83 | {\ |
| 84 | int a0, a1;\ |
| 85 | a0 = ptr[k];\ |
| 86 | a1 = ptr[8 + k];\ |
| 87 | ptr[k] = a0 + a1;\ |
| 88 | ptr[8 + k] = a0 - a1;\ |
| 89 | } |
| 90 | |
| 91 | /* only used by DV codec. The input must be interlaced. 128 is added |
| 92 | to the pixels before clamping to avoid systematic error |
| 93 | (1024*sqrt(2)) offset would be needed otherwise. */ |
| 94 | /* XXX: I think a 1.0/sqrt(2) normalization should be needed to |
| 95 | compensate the extra butterfly stage - I don't have the full DV |
| 96 | specification */ |
| 97 | void ff_simple_idct248_put(uint8_t *dest, ptrdiff_t line_size, int16_t *block) |
| 98 | { |
| 99 | int i; |
| 100 | int16_t *ptr; |
| 101 | |
| 102 | /* butterfly */ |
| 103 | ptr = block; |
| 104 | for(i=0;i<4;i++) { |
| 105 | BF(0); |
| 106 | BF(1); |
| 107 | BF(2); |
| 108 | BF(3); |
| 109 | BF(4); |
| 110 | BF(5); |
| 111 | BF(6); |
| 112 | BF(7); |
| 113 | ptr += 2 * 8; |
| 114 | } |
| 115 | |
| 116 | /* IDCT8 on each line */ |
| 117 | for(i=0; i<8; i++) { |
| 118 | idctRowCondDC_8(block + i*8, 0); |
| 119 | } |
| 120 | |
| 121 | /* IDCT4 and store */ |
| 122 | for(i=0;i<8;i++) { |
| 123 | idct4col_put(dest + i, 2 * line_size, block + i); |
| 124 | idct4col_put(dest + line_size + i, 2 * line_size, block + 8 + i); |
| 125 | } |
| 126 | } |
| 127 | |
| 128 | /* 8x4 & 4x8 WMV2 IDCT */ |
| 129 | #undef CN_SHIFT |
| 130 | #undef C_SHIFT |
| 131 | #undef C_FIX |
| 132 | #undef C1 |
| 133 | #undef C2 |
| 134 | #define CN_SHIFT 12 |
| 135 | #define C_FIX(x) ((int)((x) * M_SQRT2 * (1 << CN_SHIFT) + 0.5)) |
| 136 | #define C1 C_FIX(0.6532814824) |
| 137 | #define C2 C_FIX(0.2705980501) |
| 138 | #define C3 C_FIX(0.5) |
| 139 | #define C_SHIFT (4+1+12) |
| 140 | static inline void idct4col_add(uint8_t *dest, ptrdiff_t line_size, const int16_t *col) |
| 141 | { |
| 142 | int c0, c1, c2, c3, a0, a1, a2, a3; |
| 143 | |
| 144 | a0 = col[8*0]; |
| 145 | a1 = col[8*1]; |
| 146 | a2 = col[8*2]; |
| 147 | a3 = col[8*3]; |
| 148 | c0 = (a0 + a2)*C3 + (1 << (C_SHIFT - 1)); |
| 149 | c2 = (a0 - a2)*C3 + (1 << (C_SHIFT - 1)); |
| 150 | c1 = a1 * C1 + a3 * C2; |
| 151 | c3 = a1 * C2 - a3 * C1; |
| 152 | dest[0] = av_clip_uint8(dest[0] + ((c0 + c1) >> C_SHIFT)); |
| 153 | dest += line_size; |
| 154 | dest[0] = av_clip_uint8(dest[0] + ((c2 + c3) >> C_SHIFT)); |
| 155 | dest += line_size; |
| 156 | dest[0] = av_clip_uint8(dest[0] + ((c2 - c3) >> C_SHIFT)); |
| 157 | dest += line_size; |
| 158 | dest[0] = av_clip_uint8(dest[0] + ((c0 - c1) >> C_SHIFT)); |
| 159 | } |
| 160 | |
| 161 | #define RN_SHIFT 15 |
| 162 | #define R_FIX(x) ((int)((x) * M_SQRT2 * (1 << RN_SHIFT) + 0.5)) |
| 163 | #define R1 R_FIX(0.6532814824) |
| 164 | #define R2 R_FIX(0.2705980501) |
| 165 | #define R3 R_FIX(0.5) |
| 166 | #define R_SHIFT 11 |
| 167 | static inline void idct4row(int16_t *row) |
| 168 | { |
| 169 | int c0, c1, c2, c3, a0, a1, a2, a3; |
| 170 | |
| 171 | a0 = row[0]; |
| 172 | a1 = row[1]; |
| 173 | a2 = row[2]; |
| 174 | a3 = row[3]; |
| 175 | c0 = (a0 + a2)*R3 + (1 << (R_SHIFT - 1)); |
| 176 | c2 = (a0 - a2)*R3 + (1 << (R_SHIFT - 1)); |
| 177 | c1 = a1 * R1 + a3 * R2; |
| 178 | c3 = a1 * R2 - a3 * R1; |
| 179 | row[0]= (c0 + c1) >> R_SHIFT; |
| 180 | row[1]= (c2 + c3) >> R_SHIFT; |
| 181 | row[2]= (c2 - c3) >> R_SHIFT; |
| 182 | row[3]= (c0 - c1) >> R_SHIFT; |
| 183 | } |
| 184 | |
| 185 | void ff_simple_idct84_add(uint8_t *dest, ptrdiff_t line_size, int16_t *block) |
| 186 | { |
| 187 | int i; |
| 188 | |
| 189 | /* IDCT8 on each line */ |
| 190 | for(i=0; i<4; i++) { |
| 191 | idctRowCondDC_8(block + i*8, 0); |
| 192 | } |
| 193 | |
| 194 | /* IDCT4 and store */ |
| 195 | for(i=0;i<8;i++) { |
| 196 | idct4col_add(dest + i, line_size, block + i); |
| 197 | } |
| 198 | } |
| 199 | |
| 200 | void ff_simple_idct48_add(uint8_t *dest, ptrdiff_t line_size, int16_t *block) |
| 201 | { |
| 202 | int i; |
| 203 | |
| 204 | /* IDCT4 on each line */ |
| 205 | for(i=0; i<8; i++) { |
| 206 | idct4row(block + i*8); |
| 207 | } |
| 208 | |
| 209 | /* IDCT8 and store */ |
| 210 | for(i=0; i<4; i++){ |
| 211 | idctSparseColAdd_8(dest + i, line_size, block + i); |
| 212 | } |
| 213 | } |
| 214 | |
| 215 | void ff_simple_idct44_add(uint8_t *dest, ptrdiff_t line_size, int16_t *block) |
| 216 | { |
| 217 | int i; |
| 218 | |
| 219 | /* IDCT4 on each line */ |
| 220 | for(i=0; i<4; i++) { |
| 221 | idct4row(block + i*8); |
| 222 | } |
| 223 | |
| 224 | /* IDCT4 and store */ |
| 225 | for(i=0; i<4; i++){ |
| 226 | idct4col_add(dest + i, line_size, block + i); |
| 227 | } |
| 228 | } |
| 229 | |
| 230 | void ff_prores_idct(int16_t *block, const int16_t *qmat) |
| 231 | { |
| 232 | int i; |
| 233 | |
| 234 | for (i = 0; i < 64; i++) |
| 235 | block[i] *= qmat[i]; |
| 236 | |
| 237 | for (i = 0; i < 8; i++) |
| 238 | idctRowCondDC_extrashift_10(block + i*8, 2); |
| 239 | |
| 240 | for (i = 0; i < 8; i++) { |
| 241 | block[i] += 8192; |
| 242 | idctSparseCol_extrashift_10(block + i); |
| 243 | } |
| 244 | } |
| 245 |