FFmpeg: libavcodec/dct32.c Source File

00001 /*
00002  * Template for the Discrete Cosine Transform for 32 samples
00003  * Copyright (c) 2001, 2002 Fabrice Bellard
00004  *
00005  * This file is part of FFmpeg.
00006  *
00007  * FFmpeg is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * FFmpeg is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with FFmpeg; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 
00022 #include "dct32.h"
00023 #include "mathops.h"
00024 
00025 #if DCT32_FLOAT
00026 #   define dct32 ff_dct32_float
00027 #   define FIXHR(x)       ((float)(x))
00028 #   define MULH3(x, y, s) ((s)*(y)*(x))
00029 #   define INTFLOAT float
00030 #else
00031 #   define dct32 ff_dct32_fixed
00032 #   define FIXHR(a)       ((int)((a) * (1LL<<32) + 0.5))
00033 #   define MULH3(x, y, s) MULH((s)*(x), y)
00034 #   define INTFLOAT int
00035 #endif
00036 
00037 
00038 /* tab[i][j] = 1.0 / (2.0 * cos(pi*(2*k+1) / 2^(6 - j))) */
00039 
00040 /* cos(i*pi/64) */
00041 
00042 #define COS0_0  FIXHR(0.50060299823519630134/2)
00043 #define COS0_1  FIXHR(0.50547095989754365998/2)
00044 #define COS0_2  FIXHR(0.51544730992262454697/2)
00045 #define COS0_3  FIXHR(0.53104259108978417447/2)
00046 #define COS0_4  FIXHR(0.55310389603444452782/2)
00047 #define COS0_5  FIXHR(0.58293496820613387367/2)
00048 #define COS0_6  FIXHR(0.62250412303566481615/2)
00049 #define COS0_7  FIXHR(0.67480834145500574602/2)
00050 #define COS0_8  FIXHR(0.74453627100229844977/2)
00051 #define COS0_9  FIXHR(0.83934964541552703873/2)
00052 #define COS0_10 FIXHR(0.97256823786196069369/2)
00053 #define COS0_11 FIXHR(1.16943993343288495515/4)
00054 #define COS0_12 FIXHR(1.48416461631416627724/4)
00055 #define COS0_13 FIXHR(2.05778100995341155085/8)
00056 #define COS0_14 FIXHR(3.40760841846871878570/8)
00057 #define COS0_15 FIXHR(10.19000812354805681150/32)
00058 
00059 #define COS1_0 FIXHR(0.50241928618815570551/2)
00060 #define COS1_1 FIXHR(0.52249861493968888062/2)
00061 #define COS1_2 FIXHR(0.56694403481635770368/2)
00062 #define COS1_3 FIXHR(0.64682178335999012954/2)
00063 #define COS1_4 FIXHR(0.78815462345125022473/2)
00064 #define COS1_5 FIXHR(1.06067768599034747134/4)
00065 #define COS1_6 FIXHR(1.72244709823833392782/4)
00066 #define COS1_7 FIXHR(5.10114861868916385802/16)
00067 
00068 #define COS2_0 FIXHR(0.50979557910415916894/2)
00069 #define COS2_1 FIXHR(0.60134488693504528054/2)
00070 #define COS2_2 FIXHR(0.89997622313641570463/2)
00071 #define COS2_3 FIXHR(2.56291544774150617881/8)
00072 
00073 #define COS3_0 FIXHR(0.54119610014619698439/2)
00074 #define COS3_1 FIXHR(1.30656296487637652785/4)
00075 
00076 #define COS4_0 FIXHR(0.70710678118654752439/2)
00077 
00078 /* butterfly operator */
00079 #define BF(a, b, c, s)\
00080 {\
00081     tmp0 = val##a + val##b;\
00082     tmp1 = val##a - val##b;\
00083     val##a = tmp0;\
00084     val##b = MULH3(tmp1, c, 1<<(s));\
00085 }
00086 
00087 #define BF0(a, b, c, s)\
00088 {\
00089     tmp0 = tab[a] + tab[b];\
00090     tmp1 = tab[a] - tab[b];\
00091     val##a = tmp0;\
00092     val##b = MULH3(tmp1, c, 1<<(s));\
00093 }
00094 
00095 #define BF1(a, b, c, d)\
00096 {\
00097     BF(a, b, COS4_0, 1);\
00098     BF(c, d,-COS4_0, 1);\
00099     val##c += val##d;\
00100 }
00101 
00102 #define BF2(a, b, c, d)\
00103 {\
00104     BF(a, b, COS4_0, 1);\
00105     BF(c, d,-COS4_0, 1);\
00106     val##c += val##d;\
00107     val##a += val##c;\
00108     val##c += val##b;\
00109     val##b += val##d;\
00110 }
00111 
00112 #define ADD(a, b) val##a += val##b
00113 
00114 /* DCT32 without 1/sqrt(2) coef zero scaling. */
00115 void dct32(INTFLOAT *out, const INTFLOAT *tab)
00116 {
00117     INTFLOAT tmp0, tmp1;
00118 
00119     INTFLOAT val0 , val1 , val2 , val3 , val4 , val5 , val6 , val7 ,
00120              val8 , val9 , val10, val11, val12, val13, val14, val15,
00121              val16, val17, val18, val19, val20, val21, val22, val23,
00122              val24, val25, val26, val27, val28, val29, val30, val31;
00123 
00124     /* pass 1 */
00125     BF0( 0, 31, COS0_0 , 1);
00126     BF0(15, 16, COS0_15, 5);
00127     /* pass 2 */
00128     BF( 0, 15, COS1_0 , 1);
00129     BF(16, 31,-COS1_0 , 1);
00130     /* pass 1 */
00131     BF0( 7, 24, COS0_7 , 1);
00132     BF0( 8, 23, COS0_8 , 1);
00133     /* pass 2 */
00134     BF( 7,  8, COS1_7 , 4);
00135     BF(23, 24,-COS1_7 , 4);
00136     /* pass 3 */
00137     BF( 0,  7, COS2_0 , 1);
00138     BF( 8, 15,-COS2_0 , 1);
00139     BF(16, 23, COS2_0 , 1);
00140     BF(24, 31,-COS2_0 , 1);
00141     /* pass 1 */
00142     BF0( 3, 28, COS0_3 , 1);
00143     BF0(12, 19, COS0_12, 2);
00144     /* pass 2 */
00145     BF( 3, 12, COS1_3 , 1);
00146     BF(19, 28,-COS1_3 , 1);
00147     /* pass 1 */
00148     BF0( 4, 27, COS0_4 , 1);
00149     BF0(11, 20, COS0_11, 2);
00150     /* pass 2 */
00151     BF( 4, 11, COS1_4 , 1);
00152     BF(20, 27,-COS1_4 , 1);
00153     /* pass 3 */
00154     BF( 3,  4, COS2_3 , 3);
00155     BF(11, 12,-COS2_3 , 3);
00156     BF(19, 20, COS2_3 , 3);
00157     BF(27, 28,-COS2_3 , 3);
00158     /* pass 4 */
00159     BF( 0,  3, COS3_0 , 1);
00160     BF( 4,  7,-COS3_0 , 1);
00161     BF( 8, 11, COS3_0 , 1);
00162     BF(12, 15,-COS3_0 , 1);
00163     BF(16, 19, COS3_0 , 1);
00164     BF(20, 23,-COS3_0 , 1);
00165     BF(24, 27, COS3_0 , 1);
00166     BF(28, 31,-COS3_0 , 1);
00167 
00168 
00169 
00170     /* pass 1 */
00171     BF0( 1, 30, COS0_1 , 1);
00172     BF0(14, 17, COS0_14, 3);
00173     /* pass 2 */
00174     BF( 1, 14, COS1_1 , 1);
00175     BF(17, 30,-COS1_1 , 1);
00176     /* pass 1 */
00177     BF0( 6, 25, COS0_6 , 1);
00178     BF0( 9, 22, COS0_9 , 1);
00179     /* pass 2 */
00180     BF( 6,  9, COS1_6 , 2);
00181     BF(22, 25,-COS1_6 , 2);
00182     /* pass 3 */
00183     BF( 1,  6, COS2_1 , 1);
00184     BF( 9, 14,-COS2_1 , 1);
00185     BF(17, 22, COS2_1 , 1);
00186     BF(25, 30,-COS2_1 , 1);
00187 
00188     /* pass 1 */
00189     BF0( 2, 29, COS0_2 , 1);
00190     BF0(13, 18, COS0_13, 3);
00191     /* pass 2 */
00192     BF( 2, 13, COS1_2 , 1);
00193     BF(18, 29,-COS1_2 , 1);
00194     /* pass 1 */
00195     BF0( 5, 26, COS0_5 , 1);
00196     BF0(10, 21, COS0_10, 1);
00197     /* pass 2 */
00198     BF( 5, 10, COS1_5 , 2);
00199     BF(21, 26,-COS1_5 , 2);
00200     /* pass 3 */
00201     BF( 2,  5, COS2_2 , 1);
00202     BF(10, 13,-COS2_2 , 1);
00203     BF(18, 21, COS2_2 , 1);
00204     BF(26, 29,-COS2_2 , 1);
00205     /* pass 4 */
00206     BF( 1,  2, COS3_1 , 2);
00207     BF( 5,  6,-COS3_1 , 2);
00208     BF( 9, 10, COS3_1 , 2);
00209     BF(13, 14,-COS3_1 , 2);
00210     BF(17, 18, COS3_1 , 2);
00211     BF(21, 22,-COS3_1 , 2);
00212     BF(25, 26, COS3_1 , 2);
00213     BF(29, 30,-COS3_1 , 2);
00214 
00215     /* pass 5 */
00216     BF1( 0,  1,  2,  3);
00217     BF2( 4,  5,  6,  7);
00218     BF1( 8,  9, 10, 11);
00219     BF2(12, 13, 14, 15);
00220     BF1(16, 17, 18, 19);
00221     BF2(20, 21, 22, 23);
00222     BF1(24, 25, 26, 27);
00223     BF2(28, 29, 30, 31);
00224 
00225     /* pass 6 */
00226 
00227     ADD( 8, 12);
00228     ADD(12, 10);
00229     ADD(10, 14);
00230     ADD(14,  9);
00231     ADD( 9, 13);
00232     ADD(13, 11);
00233     ADD(11, 15);
00234 
00235     out[ 0] = val0;
00236     out[16] = val1;
00237     out[ 8] = val2;
00238     out[24] = val3;
00239     out[ 4] = val4;
00240     out[20] = val5;
00241     out[12] = val6;
00242     out[28] = val7;
00243     out[ 2] = val8;
00244     out[18] = val9;
00245     out[10] = val10;
00246     out[26] = val11;
00247     out[ 6] = val12;
00248     out[22] = val13;
00249     out[14] = val14;
00250     out[30] = val15;
00251 
00252     ADD(24, 28);
00253     ADD(28, 26);
00254     ADD(26, 30);
00255     ADD(30, 25);
00256     ADD(25, 29);
00257     ADD(29, 27);
00258     ADD(27, 31);
00259 
00260     out[ 1] = val16 + val24;
00261     out[17] = val17 + val25;
00262     out[ 9] = val18 + val26;
00263     out[25] = val19 + val27;
00264     out[ 5] = val20 + val28;
00265     out[21] = val21 + val29;
00266     out[13] = val22 + val30;
00267     out[29] = val23 + val31;
00268     out[ 3] = val24 + val20;
00269     out[19] = val25 + val21;
00270     out[11] = val26 + val22;
00271     out[27] = val27 + val23;
00272     out[ 7] = val28 + val18;
00273     out[23] = val29 + val19;
00274     out[15] = val30 + val17;
00275     out[31] = val31;
00276 }