libfaad2/fixed.h

   1 /*
   2 ** FAAD2 - Freeware Advanced Audio (AAC) Decoder including SBR decoding
   3 ** Copyright (C) 2003-2004 M. Bakker, Ahead Software AG, http://www.nero.com
   4 **
   5 ** This program is free software; you can redistribute it and/or modify
   6 ** it under the terms of the GNU General Public License as published by
   7 ** the Free Software Foundation; either version 2 of the License, or
   8 ** (at your option) any later version.
   9 **
  10 ** This program is distributed in the hope that it will be useful,
  11 ** but WITHOUT ANY WARRANTY; without even the implied warranty of
  12 ** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  13 ** GNU General Public License for more details.
  14 **
  15 ** You should have received a copy of the GNU General Public License
  16 ** along with this program; if not, write to the Free Software
  17 ** Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
  18 **
  19 ** Any non-GPL usage of this software or parts of this software is strictly
  20 ** forbidden.
  21 **
  22 ** Commercial non-GPL licensing of this software is possible.
  23 ** For more info contact Ahead Software through Mpeg4AAClicense@nero.com.
  24 **
  25 ** $Id: fixed.h,v 1.26 2004/09/04 14:56:28 menno Exp $
  26 **/
  27
  28 #ifndef __FIXED_H__
  29 #define __FIXED_H__
  30
  31 #ifdef __cplusplus
  32 extern "C" {
  33 #endif
  34
  35 #if defined(_WIN32_WCE) && defined(_ARM_)
  36 #include <cmnintrin.h>
  37 #endif
  38
  39 #define COEF_BITS 28
  40 #define COEF_PRECISION (1 << COEF_BITS)
  41 #define REAL_BITS 14 // MAXIMUM OF 14 FOR FIXED POINT SBR
  42 #define REAL_PRECISION (1 << REAL_BITS)
  43
  44 /* FRAC is the fractional only part of the fixed point number [0.0..1.0) */
  45 #define FRAC_SIZE 32 /* frac is a 32 bit integer */
  46 #define FRAC_BITS 31
  47 #define FRAC_PRECISION ((uint32_t)(1 << FRAC_BITS))
  48 #define FRAC_MAX 0x7FFFFFFF
  49
  50 typedef int32_t real_t;
  51
  52
  53 #define REAL_CONST(A) (((A) >= 0) ? ((real_t)((A)*(REAL_PRECISION)+0.5)) : ((real_t)((A)*(REAL_PRECISION)-0.5)))
  54 #define COEF_CONST(A) (((A) >= 0) ? ((real_t)((A)*(COEF_PRECISION)+0.5)) : ((real_t)((A)*(COEF_PRECISION)-0.5)))
  55 #define FRAC_CONST(A) (((A) == 1.00) ? ((real_t)FRAC_MAX) : (((A) >= 0) ? ((real_t)((A)*(FRAC_PRECISION)+0.5)) : ((real_t)((A)*(FRAC_PRECISION)-0.5))))
  56 //#define FRAC_CONST(A) (((A) >= 0) ? ((real_t)((A)*(FRAC_PRECISION)+0.5)) : ((real_t)((A)*(FRAC_PRECISION)-0.5)))
  57
  58 #define Q2_BITS 22
  59 #define Q2_PRECISION (1 << Q2_BITS)
  60 #define Q2_CONST(A) (((A) >= 0) ? ((real_t)((A)*(Q2_PRECISION)+0.5)) : ((real_t)((A)*(Q2_PRECISION)-0.5)))
  61
  62 #if defined(_WIN32) && !defined(_WIN32_WCE)
  63
  64 /* multiply with real shift */
  65 static INLINE real_t MUL_R(real_t A, real_t B)
  66 {
  67     _asm {
  68         mov eax,A
  69         imul B
  70         shrd eax,edx,REAL_BITS
  71     }
  72 }
  73
  74 /* multiply with coef shift */
  75 static INLINE real_t MUL_C(real_t A, real_t B)
  76 {
  77     _asm {
  78         mov eax,A
  79         imul B
  80         shrd eax,edx,COEF_BITS
  81     }
  82 }
  83
  84 static INLINE real_t MUL_Q2(real_t A, real_t B)
  85 {
  86     _asm {
  87         mov eax,A
  88         imul B
  89         shrd eax,edx,Q2_BITS
  90     }
  91 }
  92
  93 static INLINE real_t MUL_SHIFT6(real_t A, real_t B)
  94 {
  95     _asm {
  96         mov eax,A
  97         imul B
  98         shrd eax,edx,6
  99     }
 100 }
 101
 102 static INLINE real_t MUL_SHIFT23(real_t A, real_t B)
 103 {
 104     _asm {
 105         mov eax,A
 106         imul B
 107         shrd eax,edx,23
 108     }
 109 }
 110
 111 #if 1
 112 static INLINE real_t _MulHigh(real_t A, real_t B)
 113 {
 114     _asm {
 115         mov eax,A
 116         imul B
 117         mov eax,edx
 118     }
 119 }
 120
 121 /* multiply with fractional shift */
 122 static INLINE real_t MUL_F(real_t A, real_t B)
 123 {
 124     return _MulHigh(A,B) << (FRAC_SIZE-FRAC_BITS);
 125 }
 126
 127 /* Complex multiplication */
 128 static INLINE void ComplexMult(real_t *y1, real_t *y2,
 129     real_t x1, real_t x2, real_t c1, real_t c2)
 130 {
 131     *y1 = (_MulHigh(x1, c1) + _MulHigh(x2, c2))<<(FRAC_SIZE-FRAC_BITS);
 132     *y2 = (_MulHigh(x2, c1) - _MulHigh(x1, c2))<<(FRAC_SIZE-FRAC_BITS);
 133 }
 134 #else
 135 static INLINE real_t MUL_F(real_t A, real_t B)
 136 {
 137     _asm {
 138         mov eax,A
 139         imul B
 140         shrd eax,edx,FRAC_BITS
 141     }
 142 }
 143
 144 /* Complex multiplication */
 145 static INLINE void ComplexMult(real_t *y1, real_t *y2,
 146     real_t x1, real_t x2, real_t c1, real_t c2)
 147 {
 148     *y1 = MUL_F(x1, c1) + MUL_F(x2, c2);
 149     *y2 = MUL_F(x2, c1) - MUL_F(x1, c2);
 150 }
 151 #endif
 152
 153 #elif defined(__GNUC__) && defined (__arm__)
 154
 155 /* taken from MAD */
 156 #define arm_mul(x, y, SCALEBITS) \
 157 ({ \
 158     uint32_t __hi; \
 159     uint32_t __lo; \
 160     uint32_t __result; \
 161     asm("smull  %0, %1, %3, %4\n\t" \
 162         "movs   %0, %0, lsr %5\n\t" \
 163         "adc    %2, %0, %1, lsl %6" \
 164         : "=&r" (__lo), "=&r" (__hi), "=r" (__result) \
 165         : "%r" (x), "r" (y), \
 166         "M" (SCALEBITS), "M" (32 - (SCALEBITS)) \
 167         : "cc"); \
 168         __result; \
 169 })
 170
 171 static INLINE real_t MUL_R(real_t A, real_t B)
 172 {
 173     return arm_mul(A, B, REAL_BITS);
 174 }
 175
 176 static INLINE real_t MUL_C(real_t A, real_t B)
 177 {
 178     return arm_mul(A, B, COEF_BITS);
 179 }
 180
 181 static INLINE real_t MUL_Q2(real_t A, real_t B)
 182 {
 183     return arm_mul(A, B, Q2_BITS);
 184 }
 185
 186 static INLINE real_t MUL_SHIFT6(real_t A, real_t B)
 187 {
 188     return arm_mul(A, B, 6);
 189 }
 190
 191 static INLINE real_t MUL_SHIFT23(real_t A, real_t B)
 192 {
 193     return arm_mul(A, B, 23);
 194 }
 195
 196 static INLINE real_t _MulHigh(real_t x, real_t y)
 197 {
 198     uint32_t __lo;
 199     uint32_t __hi;
 200     asm("smull\t%0, %1, %2, %3"
 201         : "=&r"(__lo),"=&r"(__hi)
 202         : "%r"(x),"r"(y)
 203         : "cc");
 204     return __hi;
 205 }
 206
 207 static INLINE real_t MUL_F(real_t A, real_t B)
 208 {
 209     return _MulHigh(A, B) << (FRAC_SIZE-FRAC_BITS);
 210 }
 211
 212 /* Complex multiplication */
 213 static INLINE void ComplexMult(real_t *y1, real_t *y2,
 214     real_t x1, real_t x2, real_t c1, real_t c2)
 215 {
 216     int32_t tmp, yt1, yt2;
 217     asm("smull %0, %1, %4, %6\n\t"
 218         "smlal %0, %1, %5, %7\n\t"
 219         "rsb   %3, %4, #0\n\t"
 220         "smull %0, %2, %5, %6\n\t"
 221         "smlal %0, %2, %3, %7"
 222         : "=&r" (tmp), "=&r" (yt1), "=&r" (yt2), "=r" (x1)
 223         : "3" (x1), "r" (x2), "r" (c1), "r" (c2)
 224         : "cc" );
 225     *y1 = yt1 << (FRAC_SIZE-FRAC_BITS);
 226     *y2 = yt2 << (FRAC_SIZE-FRAC_BITS);
 227 }
 228
 229 #else
 230
 231   /* multiply with real shift */
 232   #define MUL_R(A,B) (real_t)(((int64_t)(A)*(int64_t)(B)+(1 << (REAL_BITS-1))) >> REAL_BITS)
 233   /* multiply with coef shift */
 234   #define MUL_C(A,B) (real_t)(((int64_t)(A)*(int64_t)(B)+(1 << (COEF_BITS-1))) >> COEF_BITS)
 235   /* multiply with fractional shift */
 236 #if defined(_WIN32_WCE) && defined(_ARM_)
 237   /* eVC for PocketPC has an intrinsic function that returns only the high 32 bits of a 32x32 bit multiply */
 238   static INLINE real_t MUL_F(real_t A, real_t B)
 239   {
 240       return _MulHigh(A,B) << (32-FRAC_BITS);
 241   }
 242 #else
 243   #define _MulHigh(A,B) (real_t)(((int64_t)(A)*(int64_t)(B)+(1 << (FRAC_SIZE-1))) >> FRAC_SIZE)
 244   #define MUL_F(A,B) (real_t)(((int64_t)(A)*(int64_t)(B)+(1 << (FRAC_BITS-1))) >> FRAC_BITS)
 245 #endif
 246   #define MUL_Q2(A,B) (real_t)(((int64_t)(A)*(int64_t)(B)+(1 << (Q2_BITS-1))) >> Q2_BITS)
 247   #define MUL_SHIFT6(A,B) (real_t)(((int64_t)(A)*(int64_t)(B)+(1 << (6-1))) >> 6)
 248   #define MUL_SHIFT23(A,B) (real_t)(((int64_t)(A)*(int64_t)(B)+(1 << (23-1))) >> 23)
 249
 250 /* Complex multiplication */
 251 static INLINE void ComplexMult(real_t *y1, real_t *y2,
 252     real_t x1, real_t x2, real_t c1, real_t c2)
 253 {
 254     *y1 = (_MulHigh(x1, c1) + _MulHigh(x2, c2))<<(FRAC_SIZE-FRAC_BITS);
 255     *y2 = (_MulHigh(x2, c1) - _MulHigh(x1, c2))<<(FRAC_SIZE-FRAC_BITS);
 256 }
 257
 258 #endif
 259
 260
 261
 262 #ifdef __cplusplus
 263 }
 264 #endif
 265 #endif