1 /* Copyright (c) 2003-2008 Jean-Marc Valin 2 Copyright (c) 2007-2008 CSIRO 3 Copyright (c) 2007-2009 Xiph.Org Foundation 4 Written by Jean-Marc Valin */ 5 /** 6 @file arch.h 7 @brief Various architecture definitions for CELT 8 */ 9 /* 10 Redistribution and use in source and binary forms, with or without 11 modification, are permitted provided that the following conditions 12 are met: 13 14 - Redistributions of source code must retain the above copyright 15 notice, this list of conditions and the following disclaimer. 16 17 - Redistributions in binary form must reproduce the above copyright 18 notice, this list of conditions and the following disclaimer in the 19 documentation and/or other materials provided with the distribution. 20 21 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 22 ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 23 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 24 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER 25 OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 26 EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 27 PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 28 PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 29 LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 30 NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 31 SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 32 */ 33 34 #ifndef ARCH_H 35 #define ARCH_H 36 37 #include "opus_types.h" 38 #include "opus_defines.h" 39 40 # if !defined(__GNUC_PREREQ) 41 # if defined(__GNUC__)&&defined(__GNUC_MINOR__) 42 # define __GNUC_PREREQ(_maj,_min) \ 43 ((__GNUC__<<16)+__GNUC_MINOR__>=((_maj)<<16)+(_min)) 44 # else 45 # define __GNUC_PREREQ(_maj,_min) 0 46 # endif 47 # endif 48 49 #define CELT_SIG_SCALE 32768.f 50 51 #define celt_fatal(str) _celt_fatal(str, __FILE__, __LINE__); 52 #ifdef ENABLE_ASSERTIONS 53 #include <stdio.h> 54 #include <stdlib.h> 55 #ifdef __GNUC__ 56 __attribute__((noreturn)) 57 #endif 58 static OPUS_INLINE void _celt_fatal(const char *str, const char *file, int line) 59 { 60 fprintf (stderr, "Fatal (internal) error in %s, line %d: %s\n", file, line, str); 61 abort(); 62 } 63 #define celt_assert(cond) {if (!(cond)) {celt_fatal("assertion failed: " #cond);}} 64 #define celt_assert2(cond, message) {if (!(cond)) {celt_fatal("assertion failed: " #cond "\n" message);}} 65 #else 66 #define celt_assert(cond) 67 #define celt_assert2(cond, message) 68 #endif 69 70 #define IMUL32(a,b) ((a)*(b)) 71 72 #define MIN16(a,b) ((a) < (b) ? (a) : (b)) /**< Minimum 16-bit value. */ 73 #define MAX16(a,b) ((a) > (b) ? (a) : (b)) /**< Maximum 16-bit value. */ 74 #define MIN32(a,b) ((a) < (b) ? (a) : (b)) /**< Minimum 32-bit value. */ 75 #define MAX32(a,b) ((a) > (b) ? (a) : (b)) /**< Maximum 32-bit value. */ 76 #define IMIN(a,b) ((a) < (b) ? (a) : (b)) /**< Minimum int value. */ 77 #define IMAX(a,b) ((a) > (b) ? (a) : (b)) /**< Maximum int value. */ 78 #define UADD32(a,b) ((a)+(b)) 79 #define USUB32(a,b) ((a)-(b)) 80 81 /* Set this if opus_int64 is a native type of the CPU. */ 82 /* Assume that all LP64 architectures have fast 64-bit types; also x86_64 83 (which can be ILP32 for x32) and Win64 (which is LLP64). */ 84 #if defined(__x86_64__) || defined(__LP64__) || defined(_WIN64) 85 #define OPUS_FAST_INT64 1 86 #else 87 #define OPUS_FAST_INT64 0 88 #endif 89 90 #define PRINT_MIPS(file) 91 92 #ifdef FIXED_POINT 93 94 typedef opus_int16 opus_val16; 95 typedef opus_int32 opus_val32; 96 97 typedef opus_val32 celt_sig; 98 typedef opus_val16 celt_norm; 99 typedef opus_val32 celt_ener; 100 101 #define Q15ONE 32767 102 103 #define SIG_SHIFT 12 104 105 #define NORM_SCALING 16384 106 107 #define DB_SHIFT 10 108 109 #define EPSILON 1 110 #define VERY_SMALL 0 111 #define VERY_LARGE16 ((opus_val16)32767) 112 #define Q15_ONE ((opus_val16)32767) 113 114 #define SCALEIN(a) (a) 115 #define SCALEOUT(a) (a) 116 117 #define ABS16(x) ((x) < 0 ? (-(x)) : (x)) 118 #define ABS32(x) ((x) < 0 ? (-(x)) : (x)) 119 120 static OPUS_INLINE opus_int16 SAT16(opus_int32 x) { 121 return x > 32767 ? 32767 : x < -32768 ? -32768 : (opus_int16)x; 122 } 123 124 #ifdef FIXED_DEBUG 125 #include "fixed_debug.h" 126 #else 127 128 #include "fixed_generic.h" 129 130 #ifdef OPUS_ARM_PRESUME_AARCH64_NEON_INTR 131 #include "arm/fixed_arm64.h" 132 #elif OPUS_ARM_INLINE_EDSP 133 #include "arm/fixed_armv5e.h" 134 #elif defined (OPUS_ARM_INLINE_ASM) 135 #include "arm/fixed_armv4.h" 136 #elif defined (BFIN_ASM) 137 #include "fixed_bfin.h" 138 #elif defined (TI_C5X_ASM) 139 #include "fixed_c5x.h" 140 #elif defined (TI_C6X_ASM) 141 #include "fixed_c6x.h" 142 #endif 143 144 #endif 145 146 #else /* FIXED_POINT */ 147 148 typedef float opus_val16; 149 typedef float opus_val32; 150 151 typedef float celt_sig; 152 typedef float celt_norm; 153 typedef float celt_ener; 154 155 #ifdef FLOAT_APPROX 156 /* This code should reliably detect NaN/inf even when -ffast-math is used. 157 Assumes IEEE 754 format. */ 158 static OPUS_INLINE int celt_isnan(float x) 159 { 160 union {float f; opus_uint32 i;} in; 161 in.f = x; 162 return ((in.i>>23)&0xFF)==0xFF && (in.i&0x007FFFFF)!=0; 163 } 164 #else 165 #ifdef __FAST_MATH__ 166 #error Cannot build libopus with -ffast-math unless FLOAT_APPROX is defined. This could result in crashes on extreme (e.g. NaN) input 167 #endif 168 #define celt_isnan(x) ((x)!=(x)) 169 #endif 170 171 #define Q15ONE 1.0f 172 173 #define NORM_SCALING 1.f 174 175 #define EPSILON 1e-15f 176 #define VERY_SMALL 1e-30f 177 #define VERY_LARGE16 1e15f 178 #define Q15_ONE ((opus_val16)1.f) 179 180 /* This appears to be the same speed as C99's fabsf() but it's more portable. */ 181 #define ABS16(x) ((float)fabs(x)) 182 #define ABS32(x) ((float)fabs(x)) 183 184 #define QCONST16(x,bits) (x) 185 #define QCONST32(x,bits) (x) 186 187 #define NEG16(x) (-(x)) 188 #define NEG32(x) (-(x)) 189 #define EXTRACT16(x) (x) 190 #define EXTEND32(x) (x) 191 #define SHR16(a,shift) (a) 192 #define SHL16(a,shift) (a) 193 #define SHR32(a,shift) (a) 194 #define SHL32(a,shift) (a) 195 #define PSHR32(a,shift) (a) 196 #define VSHR32(a,shift) (a) 197 198 #define PSHR(a,shift) (a) 199 #define SHR(a,shift) (a) 200 #define SHL(a,shift) (a) 201 #define SATURATE(x,a) (x) 202 #define SATURATE16(x) (x) 203 204 #define ROUND16(a,shift) (a) 205 #define HALF16(x) (.5f*(x)) 206 #define HALF32(x) (.5f*(x)) 207 208 #define ADD16(a,b) ((a)+(b)) 209 #define SUB16(a,b) ((a)-(b)) 210 #define ADD32(a,b) ((a)+(b)) 211 #define SUB32(a,b) ((a)-(b)) 212 #define MULT16_16_16(a,b) ((a)*(b)) 213 #define MULT16_16(a,b) ((opus_val32)(a)*(opus_val32)(b)) 214 #define MAC16_16(c,a,b) ((c)+(opus_val32)(a)*(opus_val32)(b)) 215 216 #define MULT16_32_Q15(a,b) ((a)*(b)) 217 #define MULT16_32_Q16(a,b) ((a)*(b)) 218 219 #define MULT32_32_Q31(a,b) ((a)*(b)) 220 221 #define MAC16_32_Q15(c,a,b) ((c)+(a)*(b)) 222 #define MAC16_32_Q16(c,a,b) ((c)+(a)*(b)) 223 224 #define MULT16_16_Q11_32(a,b) ((a)*(b)) 225 #define MULT16_16_Q11(a,b) ((a)*(b)) 226 #define MULT16_16_Q13(a,b) ((a)*(b)) 227 #define MULT16_16_Q14(a,b) ((a)*(b)) 228 #define MULT16_16_Q15(a,b) ((a)*(b)) 229 #define MULT16_16_P15(a,b) ((a)*(b)) 230 #define MULT16_16_P13(a,b) ((a)*(b)) 231 #define MULT16_16_P14(a,b) ((a)*(b)) 232 #define MULT16_32_P16(a,b) ((a)*(b)) 233 234 #define DIV32_16(a,b) (((opus_val32)(a))/(opus_val16)(b)) 235 #define DIV32(a,b) (((opus_val32)(a))/(opus_val32)(b)) 236 237 #define SCALEIN(a) ((a)*CELT_SIG_SCALE) 238 #define SCALEOUT(a) ((a)*(1/CELT_SIG_SCALE)) 239 240 #define SIG2WORD16(x) (x) 241 242 #endif /* !FIXED_POINT */ 243 244 #ifndef GLOBAL_STACK_SIZE 245 #ifdef FIXED_POINT 246 #define GLOBAL_STACK_SIZE 100000 247 #else 248 #define GLOBAL_STACK_SIZE 100000 249 #endif 250 #endif 251 252 #endif /* ARCH_H */ 253
