1.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     1.2 +++ b/libavcodec/simple_idct.c	Tue Sep 25 15:55:33 2012 +0200
     1.3 @@ -0,0 +1,372 @@
     1.4 +/*
     1.5 + * Simple IDCT
     1.6 + *
     1.7 + * Copyright (c) 2001 Michael Niedermayer <michaelni@gmx.at>
     1.8 + *
     1.9 + * This file is part of FFmpeg.
    1.10 + *
    1.11 + * FFmpeg is free software; you can redistribute it and/or
    1.12 + * modify it under the terms of the GNU Lesser General Public
    1.13 + * License as published by the Free Software Foundation; either
    1.14 + * version 2.1 of the License, or (at your option) any later version.
    1.15 + *
    1.16 + * FFmpeg is distributed in the hope that it will be useful,
    1.17 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
    1.18 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
    1.19 + * Lesser General Public License for more details.
    1.20 + *
    1.21 + * You should have received a copy of the GNU Lesser General Public
    1.22 + * License along with FFmpeg; if not, write to the Free Software
    1.23 + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
    1.24 + */
    1.25 +
    1.26 +/**
    1.27 + * @file
    1.28 + * simpleidct in C.
    1.29 + */
    1.30 +
    1.31 +/*
    1.32 +  based upon some outcommented c code from mpeg2dec (idct_mmx.c
    1.33 +  written by Aaron Holtzman <aholtzma@ess.engr.uvic.ca>)
    1.34 + */
    1.35 +#include "avcodec.h"
    1.36 +#include "dsputil.h"
    1.37 +#include "mathops.h"
    1.38 +#include "simple_idct.h"
    1.39 +
    1.40 +#if 0
    1.41 +#define W1 2841 /* 2048*sqrt (2)*cos (1*pi/16) */
    1.42 +#define W2 2676 /* 2048*sqrt (2)*cos (2*pi/16) */
    1.43 +#define W3 2408 /* 2048*sqrt (2)*cos (3*pi/16) */
    1.44 +#define W4 2048 /* 2048*sqrt (2)*cos (4*pi/16) */
    1.45 +#define W5 1609 /* 2048*sqrt (2)*cos (5*pi/16) */
    1.46 +#define W6 1108 /* 2048*sqrt (2)*cos (6*pi/16) */
    1.47 +#define W7 565  /* 2048*sqrt (2)*cos (7*pi/16) */
    1.48 +#define ROW_SHIFT 8
    1.49 +#define COL_SHIFT 17
    1.50 +#else
    1.51 +#define W1  22725  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
    1.52 +#define W2  21407  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
    1.53 +#define W3  19266  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
    1.54 +#define W4  16383  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
    1.55 +#define W5  12873  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
    1.56 +#define W6  8867   //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
    1.57 +#define W7  4520   //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
    1.58 +#define ROW_SHIFT 11
    1.59 +#define COL_SHIFT 20 // 6
    1.60 +#endif
    1.61 +
    1.62 +static inline void idctRowCondDC (DCTELEM * row)
    1.63 +{
    1.64 +        int a0, a1, a2, a3, b0, b1, b2, b3;
    1.65 +        uint64_t temp;
    1.66 +
    1.67 +#if HAVE_BIGENDIAN
    1.68 +#define ROW0_MASK 0xffff000000000000LL
    1.69 +#else
    1.70 +#define ROW0_MASK 0xffffLL
    1.71 +#endif
    1.72 +        if(sizeof(DCTELEM)==2){
    1.73 +            if ( ((((uint64_t *)row)[0] & ~ROW0_MASK) |
    1.74 +                  ((uint64_t *)row)[1]) == 0) {
    1.75 +                temp = (row[0] << 3) & 0xffff;
    1.76 +                temp += temp << 16;
    1.77 +                temp += temp << 32;
    1.78 +                ((uint64_t *)row)[0] = temp;
    1.79 +                ((uint64_t *)row)[1] = temp;
    1.80 +                return;
    1.81 +            }
    1.82 +        }else{
    1.83 +            if (!(row[1]|row[2]|row[3]|row[4]|row[5]|row[6]|row[7])) {
    1.84 +                row[0]=row[1]=row[2]=row[3]=row[4]=row[5]=row[6]=row[7]= row[0] << 3;
    1.85 +                return;
    1.86 +            }
    1.87 +        }
    1.88 +
    1.89 +        a0 = (W4 * row[0]) + (1 << (ROW_SHIFT - 1));
    1.90 +        a1 = a0;
    1.91 +        a2 = a0;
    1.92 +        a3 = a0;
    1.93 +
    1.94 +        /* no need to optimize : gcc does it */
    1.95 +        a0 += W2 * row[2];
    1.96 +        a1 += W6 * row[2];
    1.97 +        a2 -= W6 * row[2];
    1.98 +        a3 -= W2 * row[2];
    1.99 +
   1.100 +        b0 = MUL16(W1, row[1]);
   1.101 +        MAC16(b0, W3, row[3]);
   1.102 +        b1 = MUL16(W3, row[1]);
   1.103 +        MAC16(b1, -W7, row[3]);
   1.104 +        b2 = MUL16(W5, row[1]);
   1.105 +        MAC16(b2, -W1, row[3]);
   1.106 +        b3 = MUL16(W7, row[1]);
   1.107 +        MAC16(b3, -W5, row[3]);
   1.108 +
   1.109 +        temp = ((uint64_t*)row)[1];
   1.110 +
   1.111 +        if (temp != 0) {
   1.112 +            a0 += W4*row[4] + W6*row[6];
   1.113 +            a1 += - W4*row[4] - W2*row[6];
   1.114 +            a2 += - W4*row[4] + W2*row[6];
   1.115 +            a3 += W4*row[4] - W6*row[6];
   1.116 +
   1.117 +            MAC16(b0, W5, row[5]);
   1.118 +            MAC16(b0, W7, row[7]);
   1.119 +
   1.120 +            MAC16(b1, -W1, row[5]);
   1.121 +            MAC16(b1, -W5, row[7]);
   1.122 +
   1.123 +            MAC16(b2, W7, row[5]);
   1.124 +            MAC16(b2, W3, row[7]);
   1.125 +
   1.126 +            MAC16(b3, W3, row[5]);
   1.127 +            MAC16(b3, -W1, row[7]);
   1.128 +        }
   1.129 +
   1.130 +        row[0] = (a0 + b0) >> ROW_SHIFT;
   1.131 +        row[7] = (a0 - b0) >> ROW_SHIFT;
   1.132 +        row[1] = (a1 + b1) >> ROW_SHIFT;
   1.133 +        row[6] = (a1 - b1) >> ROW_SHIFT;
   1.134 +        row[2] = (a2 + b2) >> ROW_SHIFT;
   1.135 +        row[5] = (a2 - b2) >> ROW_SHIFT;
   1.136 +        row[3] = (a3 + b3) >> ROW_SHIFT;
   1.137 +        row[4] = (a3 - b3) >> ROW_SHIFT;
   1.138 +}
   1.139 +
   1.140 +static inline void idctSparseColPut (uint8_t *dest, int line_size,
   1.141 +                                     DCTELEM * col)
   1.142 +{
   1.143 +        int a0, a1, a2, a3, b0, b1, b2, b3;
   1.144 +        uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
   1.145 +
   1.146 +        /* XXX: I did that only to give same values as previous code */
   1.147 +        a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4));
   1.148 +        a1 = a0;
   1.149 +        a2 = a0;
   1.150 +        a3 = a0;
   1.151 +
   1.152 +        a0 +=  + W2*col[8*2];
   1.153 +        a1 +=  + W6*col[8*2];
   1.154 +        a2 +=  - W6*col[8*2];
   1.155 +        a3 +=  - W2*col[8*2];
   1.156 +
   1.157 +        b0 = MUL16(W1, col[8*1]);
   1.158 +        b1 = MUL16(W3, col[8*1]);
   1.159 +        b2 = MUL16(W5, col[8*1]);
   1.160 +        b3 = MUL16(W7, col[8*1]);
   1.161 +
   1.162 +        MAC16(b0, + W3, col[8*3]);
   1.163 +        MAC16(b1, - W7, col[8*3]);
   1.164 +        MAC16(b2, - W1, col[8*3]);
   1.165 +        MAC16(b3, - W5, col[8*3]);
   1.166 +
   1.167 +        if(col[8*4]){
   1.168 +            a0 += + W4*col[8*4];
   1.169 +            a1 += - W4*col[8*4];
   1.170 +            a2 += - W4*col[8*4];
   1.171 +            a3 += + W4*col[8*4];
   1.172 +        }
   1.173 +
   1.174 +        if (col[8*5]) {
   1.175 +            MAC16(b0, + W5, col[8*5]);
   1.176 +            MAC16(b1, - W1, col[8*5]);
   1.177 +            MAC16(b2, + W7, col[8*5]);
   1.178 +            MAC16(b3, + W3, col[8*5]);
   1.179 +        }
   1.180 +
   1.181 +        if(col[8*6]){
   1.182 +            a0 += + W6*col[8*6];
   1.183 +            a1 += - W2*col[8*6];
   1.184 +            a2 += + W2*col[8*6];
   1.185 +            a3 += - W6*col[8*6];
   1.186 +        }
   1.187 +
   1.188 +        if (col[8*7]) {
   1.189 +            MAC16(b0, + W7, col[8*7]);
   1.190 +            MAC16(b1, - W5, col[8*7]);
   1.191 +            MAC16(b2, + W3, col[8*7]);
   1.192 +            MAC16(b3, - W1, col[8*7]);
   1.193 +        }
   1.194 +
   1.195 +        dest[0] = cm[(a0 + b0) >> COL_SHIFT];
   1.196 +        dest += line_size;
   1.197 +        dest[0] = cm[(a1 + b1) >> COL_SHIFT];
   1.198 +        dest += line_size;
   1.199 +        dest[0] = cm[(a2 + b2) >> COL_SHIFT];
   1.200 +        dest += line_size;
   1.201 +        dest[0] = cm[(a3 + b3) >> COL_SHIFT];
   1.202 +        dest += line_size;
   1.203 +        dest[0] = cm[(a3 - b3) >> COL_SHIFT];
   1.204 +        dest += line_size;
   1.205 +        dest[0] = cm[(a2 - b2) >> COL_SHIFT];
   1.206 +        dest += line_size;
   1.207 +        dest[0] = cm[(a1 - b1) >> COL_SHIFT];
   1.208 +        dest += line_size;
   1.209 +        dest[0] = cm[(a0 - b0) >> COL_SHIFT];
   1.210 +}
   1.211 +
   1.212 +static inline void idctSparseColAdd (uint8_t *dest, int line_size,
   1.213 +                                     DCTELEM * col)
   1.214 +{
   1.215 +        int a0, a1, a2, a3, b0, b1, b2, b3;
   1.216 +        uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
   1.217 +
   1.218 +        /* XXX: I did that only to give same values as previous code */
   1.219 +        a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4));
   1.220 +        a1 = a0;
   1.221 +        a2 = a0;
   1.222 +        a3 = a0;
   1.223 +
   1.224 +        a0 +=  + W2*col[8*2];
   1.225 +        a1 +=  + W6*col[8*2];
   1.226 +        a2 +=  - W6*col[8*2];
   1.227 +        a3 +=  - W2*col[8*2];
   1.228 +
   1.229 +        b0 = MUL16(W1, col[8*1]);
   1.230 +        b1 = MUL16(W3, col[8*1]);
   1.231 +        b2 = MUL16(W5, col[8*1]);
   1.232 +        b3 = MUL16(W7, col[8*1]);
   1.233 +
   1.234 +        MAC16(b0, + W3, col[8*3]);
   1.235 +        MAC16(b1, - W7, col[8*3]);
   1.236 +        MAC16(b2, - W1, col[8*3]);
   1.237 +        MAC16(b3, - W5, col[8*3]);
   1.238 +
   1.239 +        if(col[8*4]){
   1.240 +            a0 += + W4*col[8*4];
   1.241 +            a1 += - W4*col[8*4];
   1.242 +            a2 += - W4*col[8*4];
   1.243 +            a3 += + W4*col[8*4];
   1.244 +        }
   1.245 +
   1.246 +        if (col[8*5]) {
   1.247 +            MAC16(b0, + W5, col[8*5]);
   1.248 +            MAC16(b1, - W1, col[8*5]);
   1.249 +            MAC16(b2, + W7, col[8*5]);
   1.250 +            MAC16(b3, + W3, col[8*5]);
   1.251 +        }
   1.252 +
   1.253 +        if(col[8*6]){
   1.254 +            a0 += + W6*col[8*6];
   1.255 +            a1 += - W2*col[8*6];
   1.256 +            a2 += + W2*col[8*6];
   1.257 +            a3 += - W6*col[8*6];
   1.258 +        }
   1.259 +
   1.260 +        if (col[8*7]) {
   1.261 +            MAC16(b0, + W7, col[8*7]);
   1.262 +            MAC16(b1, - W5, col[8*7]);
   1.263 +            MAC16(b2, + W3, col[8*7]);
   1.264 +            MAC16(b3, - W1, col[8*7]);
   1.265 +        }
   1.266 +
   1.267 +        dest[0] = cm[dest[0] + ((a0 + b0) >> COL_SHIFT)];
   1.268 +        dest += line_size;
   1.269 +        dest[0] = cm[dest[0] + ((a1 + b1) >> COL_SHIFT)];
   1.270 +        dest += line_size;
   1.271 +        dest[0] = cm[dest[0] + ((a2 + b2) >> COL_SHIFT)];
   1.272 +        dest += line_size;
   1.273 +        dest[0] = cm[dest[0] + ((a3 + b3) >> COL_SHIFT)];
   1.274 +        dest += line_size;
   1.275 +        dest[0] = cm[dest[0] + ((a3 - b3) >> COL_SHIFT)];
   1.276 +        dest += line_size;
   1.277 +        dest[0] = cm[dest[0] + ((a2 - b2) >> COL_SHIFT)];
   1.278 +        dest += line_size;
   1.279 +        dest[0] = cm[dest[0] + ((a1 - b1) >> COL_SHIFT)];
   1.280 +        dest += line_size;
   1.281 +        dest[0] = cm[dest[0] + ((a0 - b0) >> COL_SHIFT)];
   1.282 +}
   1.283 +
   1.284 +static inline void idctSparseCol (DCTELEM * col)
   1.285 +{
   1.286 +        int a0, a1, a2, a3, b0, b1, b2, b3;
   1.287 +
   1.288 +        /* XXX: I did that only to give same values as previous code */
   1.289 +        a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4));
   1.290 +        a1 = a0;
   1.291 +        a2 = a0;
   1.292 +        a3 = a0;
   1.293 +
   1.294 +        a0 +=  + W2*col[8*2];
   1.295 +        a1 +=  + W6*col[8*2];
   1.296 +        a2 +=  - W6*col[8*2];
   1.297 +        a3 +=  - W2*col[8*2];
   1.298 +
   1.299 +        b0 = MUL16(W1, col[8*1]);
   1.300 +        b1 = MUL16(W3, col[8*1]);
   1.301 +        b2 = MUL16(W5, col[8*1]);
   1.302 +        b3 = MUL16(W7, col[8*1]);
   1.303 +
   1.304 +        MAC16(b0, + W3, col[8*3]);
   1.305 +        MAC16(b1, - W7, col[8*3]);
   1.306 +        MAC16(b2, - W1, col[8*3]);
   1.307 +        MAC16(b3, - W5, col[8*3]);
   1.308 +
   1.309 +        if(col[8*4]){
   1.310 +            a0 += + W4*col[8*4];
   1.311 +            a1 += - W4*col[8*4];
   1.312 +            a2 += - W4*col[8*4];
   1.313 +            a3 += + W4*col[8*4];
   1.314 +        }
   1.315 +
   1.316 +        if (col[8*5]) {
   1.317 +            MAC16(b0, + W5, col[8*5]);
   1.318 +            MAC16(b1, - W1, col[8*5]);
   1.319 +            MAC16(b2, + W7, col[8*5]);
   1.320 +            MAC16(b3, + W3, col[8*5]);
   1.321 +        }
   1.322 +
   1.323 +        if(col[8*6]){
   1.324 +            a0 += + W6*col[8*6];
   1.325 +            a1 += - W2*col[8*6];
   1.326 +            a2 += + W2*col[8*6];
   1.327 +            a3 += - W6*col[8*6];
   1.328 +        }
   1.329 +
   1.330 +        if (col[8*7]) {
   1.331 +            MAC16(b0, + W7, col[8*7]);
   1.332 +            MAC16(b1, - W5, col[8*7]);
   1.333 +            MAC16(b2, + W3, col[8*7]);
   1.334 +            MAC16(b3, - W1, col[8*7]);
   1.335 +        }
   1.336 +
   1.337 +        col[0 ] = ((a0 + b0) >> COL_SHIFT);
   1.338 +        col[8 ] = ((a1 + b1) >> COL_SHIFT);
   1.339 +        col[16] = ((a2 + b2) >> COL_SHIFT);
   1.340 +        col[24] = ((a3 + b3) >> COL_SHIFT);
   1.341 +        col[32] = ((a3 - b3) >> COL_SHIFT);
   1.342 +        col[40] = ((a2 - b2) >> COL_SHIFT);
   1.343 +        col[48] = ((a1 - b1) >> COL_SHIFT);
   1.344 +        col[56] = ((a0 - b0) >> COL_SHIFT);
   1.345 +}
   1.346 +
   1.347 +void ff_simple_idct_put(uint8_t *dest, int line_size, DCTELEM *block)
   1.348 +{
   1.349 +    int i;
   1.350 +    for(i=0; i<8; i++)
   1.351 +        idctRowCondDC(block + i*8);
   1.352 +
   1.353 +    for(i=0; i<8; i++)
   1.354 +        idctSparseColPut(dest + i, line_size, block + i);
   1.355 +}
   1.356 +
   1.357 +void ff_simple_idct_add(uint8_t *dest, int line_size, DCTELEM *block)
   1.358 +{
   1.359 +    int i;
   1.360 +    for(i=0; i<8; i++)
   1.361 +        idctRowCondDC(block + i*8);
   1.362 +
   1.363 +    for(i=0; i<8; i++)
   1.364 +        idctSparseColAdd(dest + i, line_size, block + i);
   1.365 +}
   1.366 +
   1.367 +void ff_simple_idct(DCTELEM *block)
   1.368 +{
   1.369 +    int i;
   1.370 +    for(i=0; i<8; i++)
   1.371 +        idctRowCondDC(block + i*8);
   1.372 +
   1.373 +    for(i=0; i<8; i++)
   1.374 +        idctSparseCol(block + i);
   1.375 +}