1// This file is part of the ustl library, an STL implementation. 2// 3// Copyright (C) 2005 by Mike Sharov <msharov@users.sourceforge.net> 4// This file is free software, distributed under the MIT License. 5// 6/// \file uutility.h 7/// 8/// \brief Utility templates. 9/// 10/// Everything in here except min(), max(), distance(), and advance() 11/// are uSTL extensions and are absent from other STL implementations. 12/// 13 14#ifndef UUTILITY_H_6A58BD296269A82A4AAAA4FD19FDB3AC 15#define UUTILITY_H_6A58BD296269A82A4AAAA4FD19FDB3AC 16 17#include "uassert.h" 18#include "utypes.h" 19 20#if PLATFORM_ANDROID 21#include <stdio.h> 22#undef CPU_HAS_MMX 23#endif 24 25namespace ustl { 26 27#ifdef __GNUC__ 28 /// Returns the number of elements in a static vector 29 #define VectorSize(v) (sizeof(v) / sizeof(*v)) 30#else 31 // Old compilers will not be able to evaluate *v on an empty vector. 32 // The tradeoff here is that VectorSize will not be able to measure arrays of local structs. 33 #define VectorSize(v) (sizeof(v) / ustl::size_of_elements(1, v)) 34#endif 35 36/// Expands into a ptr,size expression for the given static vector; useful as link arguments. 37#define VectorBlock(v) (v)+0, VectorSize(v) // +0 makes it work under gcc 2.95 38/// Expands into a begin,end expression for the given static vector; useful for algorithm arguments. 39#define VectorRange(v) VectorBlock(v)+(v) 40 41/// Returns the number of bits in the given type 42#define BitsInType(t) (sizeof(t) * CHAR_BIT) 43 44/// Returns the mask of type \p t with the lowest \p n bits set. 45#define BitMask(t,n) (t(~t(0)) >> ((sizeof(t) * CHAR_BIT) - (n))) 46 47/// Argument that is used only in debug builds (as in an assert) 48#ifndef NDEBUG 49 #define DebugArg(x) x 50#else 51 #define DebugArg(x) 52#endif 53 54/// Shorthand for container iteration. 55#define foreach(type,i,ctr) for (type i = (ctr).begin(); i != (ctr).end(); ++ i) 56/// Shorthand for container reverse iteration. 57#define eachfor(type,i,ctr) for (type i = (ctr).rbegin(); i != (ctr).rend(); ++ i) 58 59/// Macro for passing template types as macro arguments. 60/// \@{ 61#define TEMPLATE_FULL_DECL1(d1,t1) template <d1 t1> 62#define TEMPLATE_FULL_DECL2(d1,t1,d2,t2) template <d1 t1, d2 t2> 63#define TEMPLATE_FULL_DECL3(d1,t1,d2,t2,d3,t3) template <d1 t1, d2 t2, d3 t3> 64#define TEMPLATE_DECL1(t1) TEMPLATE_FULL_DECL1(typename,t1) 65#define TEMPLATE_DECL2(t1,t2) TEMPLATE_FULL_DECL2(typename,t1,typename,t2) 66#define TEMPLATE_DECL3(t1,t2,t3) TEMPLATE_FULL_DECL3(typename,t1,typename,t2,typename,t3) 67#define TEMPLATE_TYPE1(type,a1) type<a1> 68#define TEMPLATE_TYPE2(type,a1,a2) type<a1,a2> 69#define TEMPLATE_TYPE3(type,a1,a2,a3) type<a1,a2,a3> 70/// \@} 71 72/// Returns the minimum of \p a and \p b 73template <typename T1, typename T2> 74inline const T1 min (const T1& a, const T2& b) 75{ 76 return (a < b ? a : b); 77} 78 79/// Returns the maximum of \p a and \p b 80template <typename T1, typename T2> 81inline const T1 max (const T1& a, const T2& b) 82{ 83 return (b < a ? a : b); 84} 85 86/// \brief Divides \p n1 by \p n2 and rounds the result up. 87/// This is in contrast to regular division, which rounds down. 88/// Negative numbers are rounded down because they are an unusual case, supporting 89/// which would require a branch. Since this is frequently used in graphics, the 90/// speed is important. 91/// 92template <typename T1, typename T2> 93inline T1 DivRU (T1 n1, T2 n2) 94{ 95 return (n1 / n2 + (n1 % n2 > 0)); 96} 97 98/// The alignment performed by default. 99const size_t c_DefaultAlignment = __alignof__(void*); 100 101/// \brief Rounds \p n up to be divisible by \p grain 102template <typename T> 103inline T Align (T n, size_t grain = c_DefaultAlignment) 104{ 105 T a, r = n % grain; 106 if (grain == 2) return (n + r); 107 switch (grain) { 108 case 4: case 8: case 16: a = (n & ~(grain - 1)) + grain; break; 109 default: a = n + (grain - r); 110 }; 111 return (r ? a : n); 112} 113 114/// Offsets an iterator 115template <typename T> 116inline T advance (T i, ssize_t offset) 117{ 118 return (i + offset); 119} 120 121#ifndef DOXYGEN_SHOULD_SKIP_THIS 122/// Offsets a void pointer 123template <> 124inline const void* advance (const void* p, ssize_t offset) 125{ 126 assert (p || !offset); 127 return (reinterpret_cast<const uint8_t*>(p) + offset); 128} 129 130/// Offsets a void pointer 131template <> 132inline void* advance (void* p, ssize_t offset) 133{ 134 assert (p || !offset); 135 return (reinterpret_cast<uint8_t*>(p) + offset); 136} 137#endif 138 139/// Returns the difference \p p1 - \p p2 140template <typename T1, typename T2> 141inline ptrdiff_t distance (T1 i1, T2 i2) 142{ 143 return (i2 - i1); 144} 145 146#ifndef DOXYGEN_SHOULD_SKIP_THIS 147#define UNVOID_DISTANCE(T1const,T2const) \ 148template <> inline ptrdiff_t distance (T1const void* p1, T2const void* p2) \ 149{ return ((T2const uint8_t*)(p2) - (T1const uint8_t*)(p1)); } 150UNVOID_DISTANCE(,) 151UNVOID_DISTANCE(const,const) 152UNVOID_DISTANCE(,const) 153UNVOID_DISTANCE(const,) 154#undef UNVOID_DISTANCE 155#endif 156 157/// \brief Returns the absolute value of \p v 158/// Unlike the stdlib functions, this is inline and works with all types. 159template <typename T> 160inline T absv (T v) 161{ 162 return (v < 0 ? -v : v); 163} 164 165/// \brief Returns -1 for negative values, 1 for positive, and 0 for 0 166template <typename T> 167inline T sign (T v) 168{ 169 return ((0 < v) - (v < 0)); 170} 171 172/// Returns the absolute value of the distance i1 and i2 173template <typename T1, typename T2> 174inline size_t abs_distance (T1 i1, T2 i2) 175{ 176 return (absv (distance(i1, i2))); 177} 178 179/// Returns the size of \p n elements of size \p T 180template <typename T> 181inline size_t size_of_elements (size_t n, const T*) 182{ 183 return (n * sizeof(T)); 184} 185 186// Defined in byteswap.h, which is usually unusable. 187#undef bswap_16 188#undef bswap_32 189#undef bswap_64 190 191#if CPU_HAS_CMPXCHG8 // If it has that, it has bswap. 192inline uint16_t bswap_16 (uint16_t v) { asm ("rorw $8, %w0" : "=r"(v) : "0"(v) : "cc"); return (v); } 193inline uint32_t bswap_32 (uint32_t v) { asm ("bswap %0" : "=r"(v) : "0"(v)); return (v); } 194#else 195inline uint16_t bswap_16 (uint16_t v) { return (v << 8 | v >> 8); } 196inline uint32_t bswap_32 (uint32_t v) { return (v << 24 | (v & 0xFF00) << 8 | (v >> 8) & 0xFF00 | v >> 24); } 197#endif 198#if HAVE_INT64_T 199inline uint64_t bswap_64 (uint64_t v) { return ((uint64_t(bswap_32(v)) << 32) | bswap_32(v >> 32)); } 200#endif 201 202/// \brief Swaps the byteorder of \p v. 203template <typename T> 204inline T bswap (const T& v) 205{ 206 switch (BitsInType(T)) { 207 default: return (v); 208 case 16: return (T (bswap_16 (uint16_t (v)))); 209 case 32: return (T (bswap_32 (uint32_t (v)))); 210#if HAVE_INT64_T 211 case 64: return (T (bswap_64 (uint64_t (v)))); 212#endif 213 }; 214} 215 216#if USTL_BYTE_ORDER == USTL_BIG_ENDIAN 217template <typename T> inline T le_to_native (const T& v) { return (bswap (v)); } 218template <typename T> inline T be_to_native (const T& v) { return (v); } 219template <typename T> inline T native_to_le (const T& v) { return (bswap (v)); } 220template <typename T> inline T native_to_be (const T& v) { return (v); } 221#elif USTL_BYTE_ORDER == USTL_LITTLE_ENDIAN 222template <typename T> inline T le_to_native (const T& v) { return (v); } 223template <typename T> inline T be_to_native (const T& v) { return (bswap (v)); } 224template <typename T> inline T native_to_le (const T& v) { return (v); } 225template <typename T> inline T native_to_be (const T& v) { return (bswap (v)); } 226#endif // USTL_BYTE_ORDER 227 228/// Deletes \p p and sets it to NULL 229template <typename T> 230inline void Delete (T*& p) 231{ 232 delete p; 233 p = NULL; 234} 235 236/// Deletes \p p as an array and sets it to NULL 237template <typename T> 238inline void DeleteVector (T*& p) 239{ 240 delete [] p; 241 p = NULL; 242} 243 244/// Template of making != from ! and == 245template <typename T> 246inline bool operator!= (const T& x, const T& y) 247{ 248 return (!(x == y)); 249} 250 251/// Template of making > from < 252template <typename T> 253inline bool operator> (const T& x, const T& y) 254{ 255 return (y < x); 256} 257 258/// Template of making <= from < and == 259template <typename T> 260inline bool operator<= (const T& x, const T& y) 261{ 262 return (!(y < x)); 263} 264 265/// Template of making >= from < and == 266template <typename T> 267inline bool operator>= (const T& x, const T& y) 268{ 269 return (!(x < y)); 270} 271 272/// Packs \p s multiple times into \p b. Useful for loop unrolling. 273template <typename TSmall, typename TBig> 274inline void pack_type (TSmall s, TBig& b) 275{ 276 const size_t n = sizeof(TBig) / sizeof(TSmall); 277 b = s; 278 // Calls to min are here to avoid warnings for shifts bigger than the type. min will be gone when optimized. 279 if (n < 2) return; 280 b = (b << min (BitsInType(TSmall), BitsInType(TBig))) | b; 281 if (n < 4) return; 282 b = (b << min (BitsInType(TSmall) * 2, BitsInType(TBig))) | b; 283 if (n < 8) return; 284 b = (b << min (BitsInType(TSmall) * 4, BitsInType(TBig))) | b; 285} 286 287#if __GNUC__ >= 3 288inline bool TestAndSet (int* pm) __attribute__((always_inline)); 289#endif 290/// Sets the contents of \p pm to 1 and returns true if the previous value was 0. 291inline bool TestAndSet (int* pm) 292{ 293#if CPU_HAS_CMPXCHG8 294 bool rv; 295 int oldVal (1); 296 asm volatile ( // cmpxchg compares to %eax and swaps if equal 297 "cmpxchgl %3, %1\n\t" 298 "sete %0" 299 : "=a" (rv), "=m" (*pm), "=r" (oldVal) 300 : "2" (oldVal), "a" (0) 301 : "memory"); 302 return (rv); 303#elif __i386__ || __x86_64__ 304 int oldVal (1); 305 asm volatile ("xchgl %0, %1" : "=r"(oldVal), "=m"(*pm) : "0"(oldVal), "m"(*pm) : "memory"); 306 return (!oldVal); 307#elif __sparc32__ // This has not been tested 308 int rv; 309 asm volatile ("ldstub %1, %0" : "=r"(rv), "=m"(*pm) : "m"(pm)); 310 return (!rv); 311#else 312 const int oldVal (*pm); 313 *pm = 1; 314 return (!oldVal); 315#endif 316} 317 318/// \brief This template is to be used for dereferencing a type-punned pointer without a warning. 319/// 320/// When casting a local variable to an unrelated type through a pointer (for 321/// example, casting a float to a uint32_t without conversion), the resulting 322/// memory location can be accessed through either pointer, which violates the 323/// strict aliasing rule. While -fno-strict-aliasing option can be given to 324/// the compiler, eliminating this warning, inefficient code may result in 325/// some instances, because aliasing inhibits some optimizations. By using 326/// this template, and by ensuring the memory is accessed in one way only, 327/// efficient code can be produced without the warning. For gcc 4.1.0+. 328/// 329template <typename DEST, typename SRC> 330inline DEST noalias (DEST, SRC* s) 331{ 332 union UPun { SRC s; DEST d; }; 333 return (((UPun*)(s))->d); 334} 335 336namespace simd { 337 /// Call after you are done using SIMD algorithms for 64 bit tuples. 338#if CPU_HAS_MMX 339 inline void reset_mmx (void) __attribute__((always_inline)); 340 #define ALL_MMX_REGS_CHANGELIST "mm0","mm1","mm2","mm3","mm4","mm5","mm6","mm7","st","st(1)","st(2)","st(3)","st(4)","st(5)","st(6)","st(7)" 341 #if CPU_HAS_3DNOW 342 inline void reset_mmx (void) { asm ("femms":::ALL_MMX_REGS_CHANGELIST); } 343 #else 344 inline void reset_mmx (void) { asm ("emms":::ALL_MMX_REGS_CHANGELIST); } 345 #endif 346#else 347 inline void reset_mmx (void) {} 348#endif 349} // namespace simd 350 351/// \brief Type that is not size_t 352/// 353/// Because size_t may be declared as unsigned long or unsigned int on 354/// different machines, this macro is convenient when defining overloads 355/// of size_t to use other types. 356/// 357#if defined(SIZE_T_IS_LONG) && !defined(__ARM_EABI__) 358 #define NOT_SIZE_T_I_OR_L unsigned int 359#else 360 #define NOT_SIZE_T_I_OR_L unsigned long 361#endif 362 363/// \brief Required when you want to overload size_t and a pointer. 364/// 365/// The compiler will happily cast a number to a pointer and declare 366/// that the overload is ambiguous unless you define overloads for all 367/// possible integral types that a number may represent. This behaviour, 368/// although braindead, is in the ANSI standard, and thus not a bug. If 369/// you want to change the standard, the best solution is to disallow any 370/// implicit casts to pointer from an integral type. Ironically, such an 371/// implicit cast is already detected by gcc. 372/// 373#if defined(USTL_ANDROID_X86) 374#define OVERLOAD_POINTER_AND_SIZE_T_V2(name, arg1type) 375#else 376#define OVERLOAD_POINTER_AND_SIZE_T_V2(name, arg1type) \ 377 inline void name (arg1type a1, short a2) { name (a1, size_t(a2)); } \ 378 inline void name (arg1type a1, unsigned short a2) { name (a1, size_t(a2)); } \ 379 inline void name (arg1type a1, int a2) { name (a1, size_t(a2)); } \ 380 inline void name (arg1type a1, long a2) { name (a1, size_t(a2)); } \ 381 inline void name (arg1type a1, NOT_SIZE_T_I_OR_L a2) { name (a1, size_t(a2)); } 382#endif 383} // namespace ustl 384 385 386#endif 387 388