ltp_arm4.h

00001 /* Copyright (C) 2004 Jean-Marc Valin */
00002 /**
00003    @file ltp_arm4.h
00004    @brief Long-Term Prediction functions (ARM4 version)
00005 */
00006 /*
00007    Redistribution and use in source and binary forms, with or without
00008    modification, are permitted provided that the following conditions
00009    are met:
00010    
00011    - Redistributions of source code must retain the above copyright
00012    notice, this list of conditions and the following disclaimer.
00013    
00014    - Redistributions in binary form must reproduce the above copyright
00015    notice, this list of conditions and the following disclaimer in the
00016    documentation and/or other materials provided with the distribution.
00017    
00018    - Neither the name of the Xiph.org Foundation nor the names of its
00019    contributors may be used to endorse or promote products derived from
00020    this software without specific prior written permission.
00021    
00022    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
00023    ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
00024    LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
00025    A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR
00026    CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
00027    EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
00028    PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
00029    PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
00030    LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
00031    NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
00032    SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
00033 */
00034 
00035 #define OVERRIDE_INNER_PROD
00036 spx_word32_t inner_prod(const spx_word16_t *x, const spx_word16_t *y, int len)
00037 {
00038    spx_word32_t sum1=0,sum2=0;
00039    spx_word16_t *deadx, *deady;
00040    int deadlen, dead1, dead2, dead3, dead4, dead5, dead6;
00041    __asm__ __volatile__ (
00042          "\tldrsh %5, [%0], #2 \n"
00043          "\tldrsh %6, [%1], #2 \n"
00044          ".inner_prod_loop%=:\n"
00045          "\tsub %7, %7, %7\n"
00046          "\tsub %10, %10, %10\n"
00047 
00048          "\tldrsh %8, [%0], #2 \n"
00049          "\tldrsh %9, [%1], #2 \n"
00050          "\tmla %7, %5, %6, %7\n"
00051          "\tldrsh %5, [%0], #2 \n"
00052          "\tldrsh %6, [%1], #2 \n"
00053          "\tmla %10, %8, %9, %10\n"
00054          "\tldrsh %8, [%0], #2 \n"
00055          "\tldrsh %9, [%1], #2 \n"
00056          "\tmla %7, %5, %6, %7\n"
00057          "\tldrsh %5, [%0], #2 \n"
00058          "\tldrsh %6, [%1], #2 \n"
00059          "\tmla %10, %8, %9, %10\n"
00060 
00061          "\tldrsh %8, [%0], #2 \n"
00062          "\tldrsh %9, [%1], #2 \n"
00063          "\tmla %7, %5, %6, %7\n"
00064          "\tldrsh %5, [%0], #2 \n"
00065          "\tldrsh %6, [%1], #2 \n"
00066          "\tmla %10, %8, %9, %10\n"
00067          "\tldrsh %8, [%0], #2 \n"
00068          "\tldrsh %9, [%1], #2 \n"
00069          "\tmla %7, %5, %6, %7\n"
00070          "\tldrsh %5, [%0], #2 \n"
00071          "\tldrsh %6, [%1], #2 \n"
00072          "\tmla %10, %8, %9, %10\n"
00073 
00074          "\tsubs %4, %4, #1\n"
00075          "\tadd %2, %2, %7, asr #5\n"
00076          "\tadd %3, %3, %10, asr #5\n"
00077          "\tbne .inner_prod_loop%=\n"
00078    : "=r" (deadx), "=r" (deady), "+r" (sum1),  "+r" (sum2),
00079      "=r" (deadlen), "=r" (dead1), "=r" (dead2), "=r" (dead3),
00080      "=r" (dead4), "=r" (dead5), "=r" (dead6)
00081    : "0" (x), "1" (y), "4" (len>>3)
00082    : "cc"
00083                         );
00084    return (sum1+sum2)>>1;
00085 }
00086 
00087 #define OVERRIDE_PITCH_XCORR
00088 void pitch_xcorr(const spx_word16_t *_x, const spx_word16_t *_y, spx_word32_t *corr, int len, int nb_pitch, char *stack)
00089 {
00090    int i,j;
00091    for (i=0;i<nb_pitch;i+=4)
00092    {
00093       /* Compute correlation*/
00094       //corr[nb_pitch-1-i]=inner_prod(x, _y+i, len);
00095       spx_word32_t sum1=0;
00096       spx_word32_t sum2=0;
00097       spx_word32_t sum3=0;
00098       spx_word32_t sum4=0;
00099       const spx_word16_t *y = _y+i;
00100       const spx_word16_t *x = _x;
00101       spx_word32_t y0, y1, y2, y3;
00102       y0=*y++;
00103       y1=*y++;
00104       y2=*y++;
00105       y3=*y++;
00106       for (j=0;j<len;j+=4)
00107       {
00108          spx_word32_t part1, part2, part3, part4, x0;
00109          spx_word32_t dead1;
00110          __asm__ __volatile__ (
00111 #ifdef SHORTCUTS
00112                "\tldrsh %10, [%8], #4 \n"
00113                "\tmul %4, %10, %0 \n"
00114                "\tldrsh %15, [%8], #4 \n"
00115                "\tmul %5, %10, %1 \n"
00116                "\tldrsh %0, [%9], #2 \n"
00117                "\tmul %6, %10, %2 \n"
00118                "\tldrsh %1, [%9], #2 \n"
00119                "\tmul %7, %10, %3 \n"
00120                
00121                
00122                "\tmla %4, %15, %2, %4 \n"
00123                "\tldrsh %2, [%9], #2 \n"
00124                "\tmla %5, %15, %3, %5 \n"
00125                "\tldrsh %3, [%9], #2 \n"
00126                "\tmla %6, %15, %0, %6 \n"
00127                "\tmla %7, %15, %1, %7 \n"
00128 
00129 #else
00130                "\tldrsh %10, [%8], #2 \n"
00131                "\tmul %4, %10, %0 \n"
00132                "\tmul %5, %10, %1 \n"
00133                "\tmul %6, %10, %2 \n"
00134                "\tmul %7, %10, %3 \n"
00135 
00136                "\tldrsh %10, [%8], #2 \n"
00137                "\tldrsh %0, [%9], #2 \n"
00138                "\tmla %4, %10, %1, %4 \n"
00139                "\tmla %5, %10, %2, %5 \n"
00140                "\tmla %6, %10, %3, %6 \n"
00141                "\tmla %7, %10, %0, %7 \n"
00142 
00143                "\tldrsh %10, [%8], #2 \n"
00144                "\tldrsh %1, [%9], #2 \n"
00145                "\tmla %4, %10, %2, %4 \n"
00146                "\tmla %5, %10, %3, %5 \n"
00147                "\tmla %6, %10, %0, %6 \n"
00148                "\tmla %7, %10, %1, %7 \n"
00149 
00150                "\tldrsh %10, [%8], #2 \n"
00151                "\tldrsh %2, [%9], #2 \n"
00152                "\tmla %4, %10, %3, %4 \n"
00153                "\tmla %5, %10, %0, %5 \n"
00154                "\tmla %6, %10, %1, %6 \n"
00155                "\tmla %7, %10, %2, %7 \n"
00156 
00157                "\tldrsh %3, [%9], #2 \n"
00158 #endif
00159 
00160                "\tldr %10, %11 \n"
00161                "\tldr %15, %12 \n"
00162                "\tadd %4, %10, %4, asr #6 \n"
00163                "\tstr %4, %11 \n"
00164                "\tldr %10, %13 \n"
00165                "\tadd %5, %15, %5, asr #6 \n"
00166                "\tstr %5, %12 \n"
00167                "\tldr %15, %14 \n"
00168                "\tadd %6, %10, %6, asr #6 \n"
00169                "\tadd %7, %15, %7, asr #6 \n"
00170                "\tstr %6, %13 \n"
00171                "\tstr %7, %14 \n"
00172 
00173             : "+r" (y0), "+r" (y1), "+r" (y2), "+r" (y3),
00174          "=r" (part1),  "=r" (part2),  "=r" (part3),  "=r" (part4),
00175               "+r" (x), "+r" (y), "=r" (x0), "+m" (sum1),
00176               "+m" (sum2), "+m" (sum3), "+m" (sum4), "=r" (dead1)
00177             :
00178             : "cc", "memory"
00179                               );
00180       }
00181       corr[nb_pitch-1-i]=sum1;
00182       corr[nb_pitch-2-i]=sum2;
00183       corr[nb_pitch-3-i]=sum3;
00184       corr[nb_pitch-4-i]=sum4;
00185    }
00186 
00187 }