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 #if OPUS_GNUC_PREREQ(3, 0) 50 #define opus_likely(x) (__builtin_expect(!!(x), 1)) 51 #define opus_unlikely(x) (__builtin_expect(!!(x), 0)) 52 #else 53 #define opus_likely(x) (!!(x)) 54 #define opus_unlikely(x) (!!(x)) 55 #endif 56 57 #define CELT_SIG_SCALE 32768.f 58 59 #define celt_fatal(str) _celt_fatal(str, __FILE__, __LINE__); 60 #ifdef ENABLE_ASSERTIONS 61 #include <stdio.h> 62 #include <stdlib.h> 63 #ifdef __GNUC__ 64 __attribute__((noreturn)) 65 #endif 66 static OPUS_INLINE void _celt_fatal(const char *str, const char *file, int line) 67 { 68 fprintf (stderr, "Fatal (internal) error in %s, line %d: %s\n", file, line, str); 69 abort(); 70 } 71 #define celt_assert(cond) {if (!(cond)) {celt_fatal("assertion failed: " #cond);}} 72 #define celt_assert2(cond, message) {if (!(cond)) {celt_fatal("assertion failed: " #cond "\n" message);}} 73 #else 74 #define celt_assert(cond) 75 #define celt_assert2(cond, message) 76 #endif 77 78 #define IMUL32(a,b) ((a)*(b)) 79 80 #define MIN16(a,b) ((a) < (b) ? (a) : (b)) /**< Minimum 16-bit value. */ 81 #define MAX16(a,b) ((a) > (b) ? (a) : (b)) /**< Maximum 16-bit value. */ 82 #define MIN32(a,b) ((a) < (b) ? (a) : (b)) /**< Minimum 32-bit value. */ 83 #define MAX32(a,b) ((a) > (b) ? (a) : (b)) /**< Maximum 32-bit value. */ 84 #define IMIN(a,b) ((a) < (b) ? (a) : (b)) /**< Minimum int value. */ 85 #define IMAX(a,b) ((a) > (b) ? (a) : (b)) /**< Maximum int value. */ 86 #define UADD32(a,b) ((a)+(b)) 87 #define USUB32(a,b) ((a)-(b)) 88 89 /* Set this if opus_int64 is a native type of the CPU. */ 90 /* Assume that all LP64 architectures have fast 64-bit types; also x86_64 91 (which can be ILP32 for x32) and Win64 (which is LLP64). */ 92 #if defined(__x86_64__) || defined(__LP64__) || defined(_WIN64) 93 #define OPUS_FAST_INT64 1 94 #else 95 #define OPUS_FAST_INT64 0 96 #endif 97 98 #define PRINT_MIPS(file) 99 100 #ifdef FIXED_POINT 101 102 typedef opus_int16 opus_val16; 103 typedef opus_int32 opus_val32; 104 typedef opus_int64 opus_val64; 105 106 typedef opus_val32 celt_sig; 107 typedef opus_val16 celt_norm; 108 typedef opus_val32 celt_ener; 109 110 #define Q15ONE 32767 111 112 #define SIG_SHIFT 12 113 /* Safe saturation value for 32-bit signals. Should be less than 114 2^31*(1-0.85) to avoid blowing up on DC at deemphasis.*/ 115 #define SIG_SAT (300000000) 116 117 #define NORM_SCALING 16384 118 119 #define DB_SHIFT 10 120 121 #define EPSILON 1 122 #define VERY_SMALL 0 123 #define VERY_LARGE16 ((opus_val16)32767) 124 #define Q15_ONE ((opus_val16)32767) 125 126 #define SCALEIN(a) (a) 127 #define SCALEOUT(a) (a) 128 129 #define ABS16(x) ((x) < 0 ? (-(x)) : (x)) 130 #define ABS32(x) ((x) < 0 ? (-(x)) : (x)) 131 132 static OPUS_INLINE opus_int16 SAT16(opus_int32 x) { 133 return x > 32767 ? 32767 : x < -32768 ? -32768 : (opus_int16)x; 134 } 135 136 #ifdef FIXED_DEBUG 137 #include "fixed_debug.h" 138 #else 139 140 #include "fixed_generic.h" 141 142 #ifdef OPUS_ARM_PRESUME_AARCH64_NEON_INTR 143 #include "arm/fixed_arm64.h" 144 #elif OPUS_ARM_INLINE_EDSP 145 #include "arm/fixed_armv5e.h" 146 #elif defined (OPUS_ARM_INLINE_ASM) 147 #include "arm/fixed_armv4.h" 148 #elif defined (BFIN_ASM) 149 #include "fixed_bfin.h" 150 #elif defined (TI_C5X_ASM) 151 #include "fixed_c5x.h" 152 #elif defined (TI_C6X_ASM) 153 #include "fixed_c6x.h" 154 #endif 155 156 #endif 157 158 #else /* FIXED_POINT */ 159 160 typedef float opus_val16; 161 typedef float opus_val32; 162 typedef float opus_val64; 163 164 typedef float celt_sig; 165 typedef float celt_norm; 166 typedef float celt_ener; 167 168 #ifdef FLOAT_APPROX 169 /* This code should reliably detect NaN/inf even when -ffast-math is used. 170 Assumes IEEE 754 format. */ 171 static OPUS_INLINE int celt_isnan(float x) 172 { 173 union {float f; opus_uint32 i;} in; 174 in.f = x; 175 return ((in.i>>23)&0xFF)==0xFF && (in.i&0x007FFFFF)!=0; 176 } 177 #else 178 #ifdef __FAST_MATH__ 179 #error Cannot build libopus with -ffast-math unless FLOAT_APPROX is defined. This could result in crashes on extreme (e.g. NaN) input 180 #endif 181 #define celt_isnan(x) ((x)!=(x)) 182 #endif 183 184 #define Q15ONE 1.0f 185 186 #define NORM_SCALING 1.f 187 188 #define EPSILON 1e-15f 189 #define VERY_SMALL 1e-30f 190 #define VERY_LARGE16 1e15f 191 #define Q15_ONE ((opus_val16)1.f) 192 193 /* This appears to be the same speed as C99's fabsf() but it's more portable. */ 194 #define ABS16(x) ((float)fabs(x)) 195 #define ABS32(x) ((float)fabs(x)) 196 197 #define QCONST16(x,bits) (x) 198 #define QCONST32(x,bits) (x) 199 200 #define NEG16(x) (-(x)) 201 #define NEG32(x) (-(x)) 202 #define NEG32_ovflw(x) (-(x)) 203 #define EXTRACT16(x) (x) 204 #define EXTEND32(x) (x) 205 #define SHR16(a,shift) (a) 206 #define SHL16(a,shift) (a) 207 #define SHR32(a,shift) (a) 208 #define SHL32(a,shift) (a) 209 #define PSHR32(a,shift) (a) 210 #define VSHR32(a,shift) (a) 211 212 #define PSHR(a,shift) (a) 213 #define SHR(a,shift) (a) 214 #define SHL(a,shift) (a) 215 #define SATURATE(x,a) (x) 216 #define SATURATE16(x) (x) 217 218 #define ROUND16(a,shift) (a) 219 #define SROUND16(a,shift) (a) 220 #define HALF16(x) (.5f*(x)) 221 #define HALF32(x) (.5f*(x)) 222 223 #define ADD16(a,b) ((a)+(b)) 224 #define SUB16(a,b) ((a)-(b)) 225 #define ADD32(a,b) ((a)+(b)) 226 #define SUB32(a,b) ((a)-(b)) 227 #define ADD32_ovflw(a,b) ((a)+(b)) 228 #define SUB32_ovflw(a,b) ((a)-(b)) 229 #define MULT16_16_16(a,b) ((a)*(b)) 230 #define MULT16_16(a,b) ((opus_val32)(a)*(opus_val32)(b)) 231 #define MAC16_16(c,a,b) ((c)+(opus_val32)(a)*(opus_val32)(b)) 232 233 #define MULT16_32_Q15(a,b) ((a)*(b)) 234 #define MULT16_32_Q16(a,b) ((a)*(b)) 235 236 #define MULT32_32_Q31(a,b) ((a)*(b)) 237 238 #define MAC16_32_Q15(c,a,b) ((c)+(a)*(b)) 239 #define MAC16_32_Q16(c,a,b) ((c)+(a)*(b)) 240 241 #define MULT16_16_Q11_32(a,b) ((a)*(b)) 242 #define MULT16_16_Q11(a,b) ((a)*(b)) 243 #define MULT16_16_Q13(a,b) ((a)*(b)) 244 #define MULT16_16_Q14(a,b) ((a)*(b)) 245 #define MULT16_16_Q15(a,b) ((a)*(b)) 246 #define MULT16_16_P15(a,b) ((a)*(b)) 247 #define MULT16_16_P13(a,b) ((a)*(b)) 248 #define MULT16_16_P14(a,b) ((a)*(b)) 249 #define MULT16_32_P16(a,b) ((a)*(b)) 250 251 #define DIV32_16(a,b) (((opus_val32)(a))/(opus_val16)(b)) 252 #define DIV32(a,b) (((opus_val32)(a))/(opus_val32)(b)) 253 254 #define SCALEIN(a) ((a)*CELT_SIG_SCALE) 255 #define SCALEOUT(a) ((a)*(1/CELT_SIG_SCALE)) 256 257 #define SIG2WORD16(x) (x) 258 259 #endif /* !FIXED_POINT */ 260 261 #ifndef GLOBAL_STACK_SIZE 262 #ifdef FIXED_POINT 263 #define GLOBAL_STACK_SIZE 120000 264 #else 265 #define GLOBAL_STACK_SIZE 120000 266 #endif 267 #endif 268 269 #endif /* ARCH_H */ 270
