ReactOS Fundraising Campaign 2012
 
€ 4,410 / € 30,000

Information | Donate

Home | Info | Community | Development | myReactOS | Contact Us

Navigation
  1. Home
  2. Community
  3. Development
  4. myReactOS
  5. Fundraiser 2012

Doxygen
  1. Main Page
  2. Alphabetical List
  3. Data Structures
  4. Directories
  5. File List
  6. Data Fields
  7. Globals
  8. Related Pages

Search

ReactOS Development > Doxygen

dct64_i386.c
Go to the documentation of this file.
00001 /*
00002     dct64_i386.c: DCT64, a C variant for i386
00003 
00004     copyright ?-2006 by the mpg123 project - free software under the terms of the LGPL 2.1
00005     see COPYING and AUTHORS files in distribution or http://mpg123.org
00006     initially written by Michael Hipp
00007 */
00008 
00009 /*
00010  * Discrete Cosine Tansform (DCT) for subband synthesis
00011  * optimized for machines with no auto-increment. 
00012  * The performance is highly compiler dependend. Maybe
00013  * the dct64.c version for 'normal' processor may be faster
00014  * even for Intel processors.
00015  */
00016 
00017 #include "mpg123lib_intern.h"
00018 
00019 static void dct64_1(real *out0,real *out1,real *b1,real *b2,real *samples)
00020 {
00021  {
00022   register real *costab = pnts[0];
00023 
00024   b1[0x00] = samples[0x00] + samples[0x1F];
00025   b1[0x01] = samples[0x01] + samples[0x1E];
00026   b1[0x1F] = REAL_MUL(samples[0x00] - samples[0x1F], costab[0x0]);
00027   b1[0x1E] = REAL_MUL(samples[0x01] - samples[0x1E], costab[0x1]);
00028 
00029   b1[0x02] = samples[0x02] + samples[0x1D];
00030   b1[0x03] = samples[0x03] + samples[0x1C];
00031   b1[0x1D] = REAL_MUL(samples[0x02] - samples[0x1D], costab[0x2]);
00032   b1[0x1C] = REAL_MUL(samples[0x03] - samples[0x1C], costab[0x3]);
00033 
00034   b1[0x04] = samples[0x04] + samples[0x1B];
00035   b1[0x05] = samples[0x05] + samples[0x1A];
00036   b1[0x1B] = REAL_MUL(samples[0x04] - samples[0x1B], costab[0x4]);
00037   b1[0x1A] = REAL_MUL(samples[0x05] - samples[0x1A], costab[0x5]);
00038 
00039   b1[0x06] = samples[0x06] + samples[0x19];
00040   b1[0x07] = samples[0x07] + samples[0x18];
00041   b1[0x19] = REAL_MUL(samples[0x06] - samples[0x19], costab[0x6]);
00042   b1[0x18] = REAL_MUL(samples[0x07] - samples[0x18], costab[0x7]);
00043 
00044   b1[0x08] = samples[0x08] + samples[0x17];
00045   b1[0x09] = samples[0x09] + samples[0x16];
00046   b1[0x17] = REAL_MUL(samples[0x08] - samples[0x17], costab[0x8]);
00047   b1[0x16] = REAL_MUL(samples[0x09] - samples[0x16], costab[0x9]);
00048 
00049   b1[0x0A] = samples[0x0A] + samples[0x15];
00050   b1[0x0B] = samples[0x0B] + samples[0x14];
00051   b1[0x15] = REAL_MUL(samples[0x0A] - samples[0x15], costab[0xA]);
00052   b1[0x14] = REAL_MUL(samples[0x0B] - samples[0x14], costab[0xB]);
00053 
00054   b1[0x0C] = samples[0x0C] + samples[0x13];
00055   b1[0x0D] = samples[0x0D] + samples[0x12];
00056   b1[0x13] = REAL_MUL(samples[0x0C] - samples[0x13], costab[0xC]);
00057   b1[0x12] = REAL_MUL(samples[0x0D] - samples[0x12], costab[0xD]);
00058 
00059   b1[0x0E] = samples[0x0E] + samples[0x11];
00060   b1[0x0F] = samples[0x0F] + samples[0x10];
00061   b1[0x11] = REAL_MUL(samples[0x0E] - samples[0x11], costab[0xE]);
00062   b1[0x10] = REAL_MUL(samples[0x0F] - samples[0x10], costab[0xF]);
00063 
00064  }
00065 
00066 
00067  {
00068   register real *costab = pnts[1];
00069 
00070   b2[0x00] = b1[0x00] + b1[0x0F]; 
00071   b2[0x01] = b1[0x01] + b1[0x0E]; 
00072   b2[0x0F] = REAL_MUL(b1[0x00] - b1[0x0F], costab[0]);
00073   b2[0x0E] = REAL_MUL(b1[0x01] - b1[0x0E], costab[1]);
00074 
00075   b2[0x02] = b1[0x02] + b1[0x0D]; 
00076   b2[0x03] = b1[0x03] + b1[0x0C]; 
00077   b2[0x0D] = REAL_MUL(b1[0x02] - b1[0x0D], costab[2]);
00078   b2[0x0C] = REAL_MUL(b1[0x03] - b1[0x0C], costab[3]);
00079 
00080   b2[0x04] = b1[0x04] + b1[0x0B]; 
00081   b2[0x05] = b1[0x05] + b1[0x0A]; 
00082   b2[0x0B] = REAL_MUL(b1[0x04] - b1[0x0B], costab[4]);
00083   b2[0x0A] = REAL_MUL(b1[0x05] - b1[0x0A], costab[5]);
00084 
00085   b2[0x06] = b1[0x06] + b1[0x09]; 
00086   b2[0x07] = b1[0x07] + b1[0x08]; 
00087   b2[0x09] = REAL_MUL(b1[0x06] - b1[0x09], costab[6]);
00088   b2[0x08] = REAL_MUL(b1[0x07] - b1[0x08], costab[7]);
00089 
00090   /* */
00091 
00092   b2[0x10] = b1[0x10] + b1[0x1F];
00093   b2[0x11] = b1[0x11] + b1[0x1E];
00094   b2[0x1F] = REAL_MUL(b1[0x1F] - b1[0x10], costab[0]);
00095   b2[0x1E] = REAL_MUL(b1[0x1E] - b1[0x11], costab[1]);
00096 
00097   b2[0x12] = b1[0x12] + b1[0x1D];
00098   b2[0x13] = b1[0x13] + b1[0x1C];
00099   b2[0x1D] = REAL_MUL(b1[0x1D] - b1[0x12], costab[2]);
00100   b2[0x1C] = REAL_MUL(b1[0x1C] - b1[0x13], costab[3]);
00101 
00102   b2[0x14] = b1[0x14] + b1[0x1B];
00103   b2[0x15] = b1[0x15] + b1[0x1A];
00104   b2[0x1B] = REAL_MUL(b1[0x1B] - b1[0x14], costab[4]);
00105   b2[0x1A] = REAL_MUL(b1[0x1A] - b1[0x15], costab[5]);
00106 
00107   b2[0x16] = b1[0x16] + b1[0x19];
00108   b2[0x17] = b1[0x17] + b1[0x18];
00109   b2[0x19] = REAL_MUL(b1[0x19] - b1[0x16], costab[6]);
00110   b2[0x18] = REAL_MUL(b1[0x18] - b1[0x17], costab[7]);
00111  }
00112 
00113  {
00114   register real *costab = pnts[2];
00115 
00116   b1[0x00] = b2[0x00] + b2[0x07];
00117   b1[0x07] = REAL_MUL(b2[0x00] - b2[0x07], costab[0]);
00118   b1[0x01] = b2[0x01] + b2[0x06];
00119   b1[0x06] = REAL_MUL(b2[0x01] - b2[0x06], costab[1]);
00120   b1[0x02] = b2[0x02] + b2[0x05];
00121   b1[0x05] = REAL_MUL(b2[0x02] - b2[0x05], costab[2]);
00122   b1[0x03] = b2[0x03] + b2[0x04];
00123   b1[0x04] = REAL_MUL(b2[0x03] - b2[0x04], costab[3]);
00124 
00125   b1[0x08] = b2[0x08] + b2[0x0F];
00126   b1[0x0F] = REAL_MUL(b2[0x0F] - b2[0x08], costab[0]);
00127   b1[0x09] = b2[0x09] + b2[0x0E];
00128   b1[0x0E] = REAL_MUL(b2[0x0E] - b2[0x09], costab[1]);
00129   b1[0x0A] = b2[0x0A] + b2[0x0D];
00130   b1[0x0D] = REAL_MUL(b2[0x0D] - b2[0x0A], costab[2]);
00131   b1[0x0B] = b2[0x0B] + b2[0x0C];
00132   b1[0x0C] = REAL_MUL(b2[0x0C] - b2[0x0B], costab[3]);
00133 
00134   b1[0x10] = b2[0x10] + b2[0x17];
00135   b1[0x17] = REAL_MUL(b2[0x10] - b2[0x17], costab[0]);
00136   b1[0x11] = b2[0x11] + b2[0x16];
00137   b1[0x16] = REAL_MUL(b2[0x11] - b2[0x16], costab[1]);
00138   b1[0x12] = b2[0x12] + b2[0x15];
00139   b1[0x15] = REAL_MUL(b2[0x12] - b2[0x15], costab[2]);
00140   b1[0x13] = b2[0x13] + b2[0x14];
00141   b1[0x14] = REAL_MUL(b2[0x13] - b2[0x14], costab[3]);
00142 
00143   b1[0x18] = b2[0x18] + b2[0x1F];
00144   b1[0x1F] = REAL_MUL(b2[0x1F] - b2[0x18], costab[0]);
00145   b1[0x19] = b2[0x19] + b2[0x1E];
00146   b1[0x1E] = REAL_MUL(b2[0x1E] - b2[0x19], costab[1]);
00147   b1[0x1A] = b2[0x1A] + b2[0x1D];
00148   b1[0x1D] = REAL_MUL(b2[0x1D] - b2[0x1A], costab[2]);
00149   b1[0x1B] = b2[0x1B] + b2[0x1C];
00150   b1[0x1C] = REAL_MUL(b2[0x1C] - b2[0x1B], costab[3]);
00151  }
00152 
00153  {
00154   register real const cos0 = pnts[3][0];
00155   register real const cos1 = pnts[3][1];
00156 
00157   b2[0x00] = b1[0x00] + b1[0x03];
00158   b2[0x03] = REAL_MUL(b1[0x00] - b1[0x03], cos0);
00159   b2[0x01] = b1[0x01] + b1[0x02];
00160   b2[0x02] = REAL_MUL(b1[0x01] - b1[0x02], cos1);
00161 
00162   b2[0x04] = b1[0x04] + b1[0x07];
00163   b2[0x07] = REAL_MUL(b1[0x07] - b1[0x04], cos0);
00164   b2[0x05] = b1[0x05] + b1[0x06];
00165   b2[0x06] = REAL_MUL(b1[0x06] - b1[0x05], cos1);
00166 
00167   b2[0x08] = b1[0x08] + b1[0x0B];
00168   b2[0x0B] = REAL_MUL(b1[0x08] - b1[0x0B], cos0);
00169   b2[0x09] = b1[0x09] + b1[0x0A];
00170   b2[0x0A] = REAL_MUL(b1[0x09] - b1[0x0A], cos1);
00171   
00172   b2[0x0C] = b1[0x0C] + b1[0x0F];
00173   b2[0x0F] = REAL_MUL(b1[0x0F] - b1[0x0C], cos0);
00174   b2[0x0D] = b1[0x0D] + b1[0x0E];
00175   b2[0x0E] = REAL_MUL(b1[0x0E] - b1[0x0D], cos1);
00176 
00177   b2[0x10] = b1[0x10] + b1[0x13];
00178   b2[0x13] = REAL_MUL(b1[0x10] - b1[0x13], cos0);
00179   b2[0x11] = b1[0x11] + b1[0x12];
00180   b2[0x12] = REAL_MUL(b1[0x11] - b1[0x12], cos1);
00181 
00182   b2[0x14] = b1[0x14] + b1[0x17];
00183   b2[0x17] = REAL_MUL(b1[0x17] - b1[0x14], cos0);
00184   b2[0x15] = b1[0x15] + b1[0x16];
00185   b2[0x16] = REAL_MUL(b1[0x16] - b1[0x15], cos1);
00186 
00187   b2[0x18] = b1[0x18] + b1[0x1B];
00188   b2[0x1B] = REAL_MUL(b1[0x18] - b1[0x1B], cos0);
00189   b2[0x19] = b1[0x19] + b1[0x1A];
00190   b2[0x1A] = REAL_MUL(b1[0x19] - b1[0x1A], cos1);
00191 
00192   b2[0x1C] = b1[0x1C] + b1[0x1F];
00193   b2[0x1F] = REAL_MUL(b1[0x1F] - b1[0x1C], cos0);
00194   b2[0x1D] = b1[0x1D] + b1[0x1E];
00195   b2[0x1E] = REAL_MUL(b1[0x1E] - b1[0x1D], cos1);
00196  }
00197 
00198  {
00199   register real const cos0 = pnts[4][0];
00200 
00201   b1[0x00] = b2[0x00] + b2[0x01];
00202   b1[0x01] = REAL_MUL(b2[0x00] - b2[0x01], cos0);
00203   b1[0x02] = b2[0x02] + b2[0x03];
00204   b1[0x03] = REAL_MUL(b2[0x03] - b2[0x02], cos0);
00205   b1[0x02] += b1[0x03];
00206 
00207   b1[0x04] = b2[0x04] + b2[0x05];
00208   b1[0x05] = REAL_MUL(b2[0x04] - b2[0x05], cos0);
00209   b1[0x06] = b2[0x06] + b2[0x07];
00210   b1[0x07] = REAL_MUL(b2[0x07] - b2[0x06], cos0);
00211   b1[0x06] += b1[0x07];
00212   b1[0x04] += b1[0x06];
00213   b1[0x06] += b1[0x05];
00214   b1[0x05] += b1[0x07];
00215 
00216   b1[0x08] = b2[0x08] + b2[0x09];
00217   b1[0x09] = REAL_MUL(b2[0x08] - b2[0x09], cos0);
00218   b1[0x0A] = b2[0x0A] + b2[0x0B];
00219   b1[0x0B] = REAL_MUL(b2[0x0B] - b2[0x0A], cos0);
00220   b1[0x0A] += b1[0x0B];
00221 
00222   b1[0x0C] = b2[0x0C] + b2[0x0D];
00223   b1[0x0D] = REAL_MUL(b2[0x0C] - b2[0x0D], cos0);
00224   b1[0x0E] = b2[0x0E] + b2[0x0F];
00225   b1[0x0F] = REAL_MUL(b2[0x0F] - b2[0x0E], cos0);
00226   b1[0x0E] += b1[0x0F];
00227   b1[0x0C] += b1[0x0E];
00228   b1[0x0E] += b1[0x0D];
00229   b1[0x0D] += b1[0x0F];
00230 
00231   b1[0x10] = b2[0x10] + b2[0x11];
00232   b1[0x11] = REAL_MUL(b2[0x10] - b2[0x11], cos0);
00233   b1[0x12] = b2[0x12] + b2[0x13];
00234   b1[0x13] = REAL_MUL(b2[0x13] - b2[0x12], cos0);
00235   b1[0x12] += b1[0x13];
00236 
00237   b1[0x14] = b2[0x14] + b2[0x15];
00238   b1[0x15] = REAL_MUL(b2[0x14] - b2[0x15], cos0);
00239   b1[0x16] = b2[0x16] + b2[0x17];
00240   b1[0x17] = REAL_MUL(b2[0x17] - b2[0x16], cos0);
00241   b1[0x16] += b1[0x17];
00242   b1[0x14] += b1[0x16];
00243   b1[0x16] += b1[0x15];
00244   b1[0x15] += b1[0x17];
00245 
00246   b1[0x18] = b2[0x18] + b2[0x19];
00247   b1[0x19] = REAL_MUL(b2[0x18] - b2[0x19], cos0);
00248   b1[0x1A] = b2[0x1A] + b2[0x1B];
00249   b1[0x1B] = REAL_MUL(b2[0x1B] - b2[0x1A], cos0);
00250   b1[0x1A] += b1[0x1B];
00251 
00252   b1[0x1C] = b2[0x1C] + b2[0x1D];
00253   b1[0x1D] = REAL_MUL(b2[0x1C] - b2[0x1D], cos0);
00254   b1[0x1E] = b2[0x1E] + b2[0x1F];
00255   b1[0x1F] = REAL_MUL(b2[0x1F] - b2[0x1E], cos0);
00256   b1[0x1E] += b1[0x1F];
00257   b1[0x1C] += b1[0x1E];
00258   b1[0x1E] += b1[0x1D];
00259   b1[0x1D] += b1[0x1F];
00260  }
00261 
00262  out0[0x10*16] = REAL_SCALE_DCT64(b1[0x00]);
00263  out0[0x10*12] = REAL_SCALE_DCT64(b1[0x04]);
00264  out0[0x10* 8] = REAL_SCALE_DCT64(b1[0x02]);
00265  out0[0x10* 4] = REAL_SCALE_DCT64(b1[0x06]);
00266  out0[0x10* 0] = REAL_SCALE_DCT64(b1[0x01]);
00267  out1[0x10* 0] = REAL_SCALE_DCT64(b1[0x01]);
00268  out1[0x10* 4] = REAL_SCALE_DCT64(b1[0x05]);
00269  out1[0x10* 8] = REAL_SCALE_DCT64(b1[0x03]);
00270  out1[0x10*12] = REAL_SCALE_DCT64(b1[0x07]);
00271 
00272 #if 1
00273  out0[0x10*14] = REAL_SCALE_DCT64(b1[0x08] + b1[0x0C]);
00274  out0[0x10*10] = REAL_SCALE_DCT64(b1[0x0C] + b1[0x0a]);
00275  out0[0x10* 6] = REAL_SCALE_DCT64(b1[0x0A] + b1[0x0E]);
00276  out0[0x10* 2] = REAL_SCALE_DCT64(b1[0x0E] + b1[0x09]);
00277  out1[0x10* 2] = REAL_SCALE_DCT64(b1[0x09] + b1[0x0D]);
00278  out1[0x10* 6] = REAL_SCALE_DCT64(b1[0x0D] + b1[0x0B]);
00279  out1[0x10*10] = REAL_SCALE_DCT64(b1[0x0B] + b1[0x0F]);
00280  out1[0x10*14] = REAL_SCALE_DCT64(b1[0x0F]);
00281 #else
00282  b1[0x08] += b1[0x0C];
00283  out0[0x10*14] = REAL_SCALE_DCT64(b1[0x08]);
00284  b1[0x0C] += b1[0x0a];
00285  out0[0x10*10] = REAL_SCALE_DCT64(b1[0x0C]);
00286  b1[0x0A] += b1[0x0E];
00287  out0[0x10* 6] = REAL_SCALE_DCT64(b1[0x0A]);
00288  b1[0x0E] += b1[0x09];
00289  out0[0x10* 2] = REAL_SCALE_DCT64(b1[0x0E]);
00290  b1[0x09] += b1[0x0D];
00291  out1[0x10* 2] = REAL_SCALE_DCT64(b1[0x09]);
00292  b1[0x0D] += b1[0x0B];
00293  out1[0x10* 6] = REAL_SCALE_DCT64(b1[0x0D]);
00294  b1[0x0B] += b1[0x0F];
00295  out1[0x10*10] = REAL_SCALE_DCT64(b1[0x0B]);
00296  out1[0x10*14] = REAL_SCALE_DCT64(b1[0x0F]);
00297 #endif
00298 
00299  { 
00300  real tmp;
00301  tmp = b1[0x18] + b1[0x1C];
00302  out0[0x10*15] = REAL_SCALE_DCT64(tmp + b1[0x10]);
00303  out0[0x10*13] = REAL_SCALE_DCT64(tmp + b1[0x14]);
00304  tmp = b1[0x1C] + b1[0x1A];
00305  out0[0x10*11] = REAL_SCALE_DCT64(tmp + b1[0x14]);
00306  out0[0x10* 9] = REAL_SCALE_DCT64(tmp + b1[0x12]);
00307  tmp = b1[0x1A] + b1[0x1E];
00308  out0[0x10* 7] = REAL_SCALE_DCT64(tmp + b1[0x12]);
00309  out0[0x10* 5] = REAL_SCALE_DCT64(tmp + b1[0x16]);
00310  tmp = b1[0x1E] + b1[0x19];
00311  out0[0x10* 3] = REAL_SCALE_DCT64(tmp + b1[0x16]);
00312  out0[0x10* 1] = REAL_SCALE_DCT64(tmp + b1[0x11]);
00313  tmp = b1[0x19] + b1[0x1D];
00314  out1[0x10* 1] = REAL_SCALE_DCT64(tmp + b1[0x11]);
00315  out1[0x10* 3] = REAL_SCALE_DCT64(tmp + b1[0x15]); 
00316  tmp = b1[0x1D] + b1[0x1B];
00317  out1[0x10* 5] = REAL_SCALE_DCT64(tmp + b1[0x15]);
00318  out1[0x10* 7] = REAL_SCALE_DCT64(tmp + b1[0x13]);
00319  tmp = b1[0x1B] + b1[0x1F];
00320  out1[0x10* 9] = REAL_SCALE_DCT64(tmp + b1[0x13]);
00321  out1[0x10*11] = REAL_SCALE_DCT64(tmp + b1[0x17]);
00322  out1[0x10*13] = REAL_SCALE_DCT64(b1[0x17] + b1[0x1F]);
00323  out1[0x10*15] = REAL_SCALE_DCT64(b1[0x1F]);
00324  }
00325 }
00326 
00327 /*
00328  * the call via dct64 is a trick to force GCC to use
00329  * (new) registers for the b1,b2 pointer to the bufs[xx] field
00330  */
00331 void dct64_i386(real *a,real *b,real *c)
00332 {
00333   real bufs[0x40];
00334   dct64_1(a,b,bufs,bufs+0x20,c);
00335 }
00336
Generated on Sat May 26 2012 04:33:00 for ReactOS by doxygen 1.7.6.1
ReactOS is a registered trademark or a trademark of ReactOS Foundation in the United States and other countries.