libavcodec/simple_idct_template.c

   1 /*
   2  * Simple IDCT
   3  *
   4  * Copyright (c) 2001 Michael Niedermayer <michaelni@gmx.at>
   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 /**
  24  * @file
  25  * simpleidct in C.
  26  */
  27
  28 /* Based upon some commented-out C code from mpeg2dec (idct_mmx.c
  29  * written by Aaron Holtzman <aholtzma@ess.engr.uvic.ca>). */
  30
  31 #include "simple_idct.h"
  32
  33 #include "bit_depth_template.c"
  34
  35 #undef W1
  36 #undef W2
  37 #undef W3
  38 #undef W4
  39 #undef W5
  40 #undef W6
  41 #undef W7
  42 #undef ROW_SHIFT
  43 #undef COL_SHIFT
  44 #undef DC_SHIFT
  45 #undef MUL
  46 #undef MAC
  47
  48 #if BIT_DEPTH == 8
  49
  50 #define W1  22725  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
  51 #define W2  21407  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
  52 #define W3  19266  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
  53 #define W4  16383  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
  54 #define W5  12873  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
  55 #define W6  8867   //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
  56 #define W7  4520   //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
  57
  58 #define ROW_SHIFT 11
  59 #define COL_SHIFT 20
  60 #define DC_SHIFT 3
  61
  62 #define MUL(a, b)    MUL16(a, b)
  63 #define MAC(a, b, c) MAC16(a, b, c)
  64
  65 #elif BIT_DEPTH == 10
  66
  67 #define W1 90901
  68 #define W2 85627
  69 #define W3 77062
  70 #define W4 65535
  71 #define W5 51491
  72 #define W6 35468
  73 #define W7 18081
  74
  75 #define ROW_SHIFT 15
  76 #define COL_SHIFT 20
  77 #define DC_SHIFT 1
  78
  79 #define MUL(a, b)    ((a) * (b))
  80 #define MAC(a, b, c) ((a) += (b) * (c))
  81
  82 #else
  83
  84 #error "Unsupported bitdepth"
  85
  86 #endif
  87
  88 static inline void FUNC(idctRowCondDC)(int16_t *row, int extra_shift)
  89 {
  90     int a0, a1, a2, a3, b0, b1, b2, b3;
  91
  92 #if HAVE_FAST_64BIT
  93 #define ROW0_MASK (0xffffLL << 48 * HAVE_BIGENDIAN)
  94     if (((AV_RN64A(row) & ~ROW0_MASK) | AV_RN64A(row+4)) == 0) {
  95         uint64_t temp;
  96         if (DC_SHIFT - extra_shift > 0) {
  97             temp = (row[0] * (1 << (DC_SHIFT - extra_shift))) & 0xffff;
  98         } else {
  99             temp = (row[0] >> (extra_shift - DC_SHIFT)) & 0xffff;
 100         }
 101         temp += temp * (1 << 16);
 102         temp += temp * ((uint64_t) 1 << 32);
 103         AV_WN64A(row, temp);
 104         AV_WN64A(row + 4, temp);
 105         return;
 106     }
 107 #else
 108     if (!(AV_RN32A(row+2) |
 109           AV_RN32A(row+4) |
 110           AV_RN32A(row+6) |
 111           row[1])) {
 112         uint32_t temp;
 113         if (DC_SHIFT - extra_shift > 0) {
 114             temp = (row[0] * (1 << (DC_SHIFT - extra_shift))) & 0xffff;
 115         } else {
 116             temp = (row[0] >> (extra_shift - DC_SHIFT)) & 0xffff;
 117         }
 118         temp += temp * (1 << 16);
 119         AV_WN32A(row, temp);
 120         AV_WN32A(row+2, temp);
 121         AV_WN32A(row+4, temp);
 122         AV_WN32A(row+6, temp);
 123         return;
 124     }
 125 #endif
 126
 127     a0 = (W4 * row[0]) + (1 << (ROW_SHIFT - 1));
 128     a1 = a0;
 129     a2 = a0;
 130     a3 = a0;
 131
 132     a0 += W2 * row[2];
 133     a1 += W6 * row[2];
 134     a2 -= W6 * row[2];
 135     a3 -= W2 * row[2];
 136
 137     b0 = MUL(W1, row[1]);
 138     MAC(b0, W3, row[3]);
 139     b1 = MUL(W3, row[1]);
 140     MAC(b1, -W7, row[3]);
 141     b2 = MUL(W5, row[1]);
 142     MAC(b2, -W1, row[3]);
 143     b3 = MUL(W7, row[1]);
 144     MAC(b3, -W5, row[3]);
 145
 146     if (AV_RN64A(row + 4)) {
 147         a0 +=   W4*row[4] + W6*row[6];
 148         a1 += - W4*row[4] - W2*row[6];
 149         a2 += - W4*row[4] + W2*row[6];
 150         a3 +=   W4*row[4] - W6*row[6];
 151
 152         MAC(b0,  W5, row[5]);
 153         MAC(b0,  W7, row[7]);
 154
 155         MAC(b1, -W1, row[5]);
 156         MAC(b1, -W5, row[7]);
 157
 158         MAC(b2,  W7, row[5]);
 159         MAC(b2,  W3, row[7]);
 160
 161         MAC(b3,  W3, row[5]);
 162         MAC(b3, -W1, row[7]);
 163     }
 164
 165     row[0] = (a0 + b0) >> (ROW_SHIFT + extra_shift);
 166     row[7] = (a0 - b0) >> (ROW_SHIFT + extra_shift);
 167     row[1] = (a1 + b1) >> (ROW_SHIFT + extra_shift);
 168     row[6] = (a1 - b1) >> (ROW_SHIFT + extra_shift);
 169     row[2] = (a2 + b2) >> (ROW_SHIFT + extra_shift);
 170     row[5] = (a2 - b2) >> (ROW_SHIFT + extra_shift);
 171     row[3] = (a3 + b3) >> (ROW_SHIFT + extra_shift);
 172     row[4] = (a3 - b3) >> (ROW_SHIFT + extra_shift);
 173 }
 174
 175 #define IDCT_COLS do {                                  \
 176         a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4)); \
 177         a1 = a0;                                        \
 178         a2 = a0;                                        \
 179         a3 = a0;                                        \
 180                                                         \
 181         a0 +=  W2*col[8*2];                             \
 182         a1 +=  W6*col[8*2];                             \
 183         a2 += -W6*col[8*2];                             \
 184         a3 += -W2*col[8*2];                             \
 185                                                         \
 186         b0 = MUL(W1, col[8*1]);                         \
 187         b1 = MUL(W3, col[8*1]);                         \
 188         b2 = MUL(W5, col[8*1]);                         \
 189         b3 = MUL(W7, col[8*1]);                         \
 190                                                         \
 191         MAC(b0,  W3, col[8*3]);                         \
 192         MAC(b1, -W7, col[8*3]);                         \
 193         MAC(b2, -W1, col[8*3]);                         \
 194         MAC(b3, -W5, col[8*3]);                         \
 195                                                         \
 196         if (col[8*4]) {                                 \
 197             a0 +=  W4*col[8*4];                         \
 198             a1 += -W4*col[8*4];                         \
 199             a2 += -W4*col[8*4];                         \
 200             a3 +=  W4*col[8*4];                         \
 201         }                                               \
 202                                                         \
 203         if (col[8*5]) {                                 \
 204             MAC(b0,  W5, col[8*5]);                     \
 205             MAC(b1, -W1, col[8*5]);                     \
 206             MAC(b2,  W7, col[8*5]);                     \
 207             MAC(b3,  W3, col[8*5]);                     \
 208         }                                               \
 209                                                         \
 210         if (col[8*6]) {                                 \
 211             a0 +=  W6*col[8*6];                         \
 212             a1 += -W2*col[8*6];                         \
 213             a2 +=  W2*col[8*6];                         \
 214             a3 += -W6*col[8*6];                         \
 215         }                                               \
 216                                                         \
 217         if (col[8*7]) {                                 \
 218             MAC(b0,  W7, col[8*7]);                     \
 219             MAC(b1, -W5, col[8*7]);                     \
 220             MAC(b2,  W3, col[8*7]);                     \
 221             MAC(b3, -W1, col[8*7]);                     \
 222         }                                               \
 223     } while (0)
 224
 225 static inline void FUNC(idctSparseColPut)(pixel *dest, ptrdiff_t line_size,
 226                                           int16_t *col)
 227 {
 228     int a0, a1, a2, a3, b0, b1, b2, b3;
 229
 230     IDCT_COLS;
 231
 232     dest[0] = av_clip_pixel((a0 + b0) >> COL_SHIFT);
 233     dest += line_size;
 234     dest[0] = av_clip_pixel((a1 + b1) >> COL_SHIFT);
 235     dest += line_size;
 236     dest[0] = av_clip_pixel((a2 + b2) >> COL_SHIFT);
 237     dest += line_size;
 238     dest[0] = av_clip_pixel((a3 + b3) >> COL_SHIFT);
 239     dest += line_size;
 240     dest[0] = av_clip_pixel((a3 - b3) >> COL_SHIFT);
 241     dest += line_size;
 242     dest[0] = av_clip_pixel((a2 - b2) >> COL_SHIFT);
 243     dest += line_size;
 244     dest[0] = av_clip_pixel((a1 - b1) >> COL_SHIFT);
 245     dest += line_size;
 246     dest[0] = av_clip_pixel((a0 - b0) >> COL_SHIFT);
 247 }
 248
 249 static inline void FUNC(idctSparseColAdd)(pixel *dest, ptrdiff_t line_size,
 250                                           int16_t *col)
 251 {
 252     int a0, a1, a2, a3, b0, b1, b2, b3;
 253
 254     IDCT_COLS;
 255
 256     dest[0] = av_clip_pixel(dest[0] + ((a0 + b0) >> COL_SHIFT));
 257     dest += line_size;
 258     dest[0] = av_clip_pixel(dest[0] + ((a1 + b1) >> COL_SHIFT));
 259     dest += line_size;
 260     dest[0] = av_clip_pixel(dest[0] + ((a2 + b2) >> COL_SHIFT));
 261     dest += line_size;
 262     dest[0] = av_clip_pixel(dest[0] + ((a3 + b3) >> COL_SHIFT));
 263     dest += line_size;
 264     dest[0] = av_clip_pixel(dest[0] + ((a3 - b3) >> COL_SHIFT));
 265     dest += line_size;
 266     dest[0] = av_clip_pixel(dest[0] + ((a2 - b2) >> COL_SHIFT));
 267     dest += line_size;
 268     dest[0] = av_clip_pixel(dest[0] + ((a1 - b1) >> COL_SHIFT));
 269     dest += line_size;
 270     dest[0] = av_clip_pixel(dest[0] + ((a0 - b0) >> COL_SHIFT));
 271 }
 272
 273 static inline void FUNC(idctSparseCol)(int16_t *col)
 274 {
 275     int a0, a1, a2, a3, b0, b1, b2, b3;
 276
 277     IDCT_COLS;
 278
 279     col[0 ] = ((a0 + b0) >> COL_SHIFT);
 280     col[8 ] = ((a1 + b1) >> COL_SHIFT);
 281     col[16] = ((a2 + b2) >> COL_SHIFT);
 282     col[24] = ((a3 + b3) >> COL_SHIFT);
 283     col[32] = ((a3 - b3) >> COL_SHIFT);
 284     col[40] = ((a2 - b2) >> COL_SHIFT);
 285     col[48] = ((a1 - b1) >> COL_SHIFT);
 286     col[56] = ((a0 - b0) >> COL_SHIFT);
 287 }
 288
 289 void FUNC(ff_simple_idct_put)(uint8_t *dest_, ptrdiff_t line_size, int16_t *block)
 290 {
 291     pixel *dest = (pixel *)dest_;
 292     int i;
 293
 294     line_size /= sizeof(pixel);
 295
 296     for (i = 0; i < 8; i++)
 297         FUNC(idctRowCondDC)(block + i*8, 0);
 298
 299     for (i = 0; i < 8; i++)
 300         FUNC(idctSparseColPut)(dest + i, line_size, block + i);
 301 }
 302
 303 void FUNC(ff_simple_idct_add)(uint8_t *dest_, ptrdiff_t line_size, int16_t *block)
 304 {
 305     pixel *dest = (pixel *)dest_;
 306     int i;
 307
 308     line_size /= sizeof(pixel);
 309
 310     for (i = 0; i < 8; i++)
 311         FUNC(idctRowCondDC)(block + i*8, 0);
 312
 313     for (i = 0; i < 8; i++)
 314         FUNC(idctSparseColAdd)(dest + i, line_size, block + i);
 315 }
 316
 317 void FUNC(ff_simple_idct)(int16_t *block)
 318 {
 319     int i;
 320
 321     for (i = 0; i < 8; i++)
 322         FUNC(idctRowCondDC)(block + i*8, 0);
 323
 324     for (i = 0; i < 8; i++)
 325         FUNC(idctSparseCol)(block + i);
 326 }