2 * Mesa 3-D graphics library
4 * Copyright (C) 1999-2008 Brian Paul All Rights Reserved.
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the "Software"),
8 * to deal in the Software without restriction, including without limitation
9 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
10 * and/or sell copies of the Software, and to permit persons to whom the
11 * Software is furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included
14 * in all copies or substantial portions of the Software.
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
17 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
20 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
21 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
22 * OTHER DEALINGS IN THE SOFTWARE.
28 * Standard C library function wrappers.
30 * This file provides wrappers for all the standard C library functions
31 * like malloc(), free(), printf(), getenv(), etc.
48 /**********************************************************************/
52 /** Allocate a structure of type \p T */
53 #define MALLOC_STRUCT(T) (struct T *) malloc(sizeof(struct T))
54 /** Allocate and zero a structure of type \p T */
55 #define CALLOC_STRUCT(T) (struct T *) calloc(1, sizeof(struct T))
61 * For GL_ARB_vertex_buffer_object we need to treat vertex array pointers
62 * as offsets into buffer stores. Since the vertex array pointer and
63 * buffer store pointer are both pointers and we need to add them, we use
65 * Both pointers/offsets are expressed in bytes.
67 #define ADD_POINTERS(A, B) ( (GLubyte *) (A) + (uintptr_t) (B) )
71 * Sometimes we treat GLfloats as GLints. On x86 systems, moving a float
72 * as a int (thereby using integer registers instead of FP registers) is
73 * a performance win. Typically, this can be done with ordinary casts.
74 * But with gcc's -fstrict-aliasing flag (which defaults to on in gcc 3.0)
75 * these casts generate warnings.
76 * The following union typedef is used to solve that.
78 typedef union { GLfloat f; GLint i; GLuint u; } fi_type;
82 /**********************************************************************
86 #define MAX_GLUSHORT 0xffff
87 #define MAX_GLUINT 0xffffffff
89 /* Degrees to radians conversion: */
90 #define DEG2RAD (M_PI/180.0)
94 * \name Work-arounds for platforms that lack C99 math functions
97 #if (!defined(_XOPEN_SOURCE) || (_XOPEN_SOURCE < 600)) && !defined(_ISOC99_SOURCE) \
98 && (!defined(__STDC_VERSION__) || (__STDC_VERSION__ < 199901L)) \
99 && (!defined(_MSC_VER) || (_MSC_VER < 1400))
100 #define acosf(f) ((float) acos(f))
101 #define asinf(f) ((float) asin(f))
102 #define atan2f(x,y) ((float) atan2(x,y))
103 #define atanf(f) ((float) atan(f))
104 #define ceilf(f) ((float) ceil(f))
105 #define cosf(f) ((float) cos(f))
106 #define coshf(f) ((float) cosh(f))
107 #define expf(f) ((float) exp(f))
108 #define exp2f(f) ((float) exp2(f))
109 #define floorf(f) ((float) floor(f))
110 #define logf(f) ((float) log(f))
113 #define log2f(f) (logf(f) * (float) (1.0 / M_LN2))
115 #define log2f(f) ((float) log2(f))
118 #define powf(x,y) ((float) pow(x,y))
119 #define sinf(f) ((float) sin(f))
120 #define sinhf(f) ((float) sinh(f))
121 #define sqrtf(f) ((float) sqrt(f))
122 #define tanf(f) ((float) tan(f))
123 #define tanhf(f) ((float) tanh(f))
124 #define acoshf(f) ((float) acosh(f))
125 #define asinhf(f) ((float) asinh(f))
126 #define atanhf(f) ((float) atanh(f))
129 #if defined(_MSC_VER)
130 #if _MSC_VER < 1800 /* Not req'd on VS2013 and above */
131 static inline float truncf(float x) { return x < 0.0f ? ceilf(x) : floorf(x); }
132 static inline float exp2f(float x) { return powf(2.0f, x); }
133 static inline float log2f(float x) { return logf(x) * 1.442695041f; }
134 static inline float asinhf(float x) { return logf(x + sqrtf(x * x + 1.0f)); }
135 static inline float acoshf(float x) { return logf(x + sqrtf(x * x - 1.0f)); }
136 static inline float atanhf(float x) { return (logf(1.0f + x) - logf(1.0f - x)) / 2.0f; }
137 static inline int isblank(int ch) { return ch == ' ' || ch == '\t'; }
138 #define strtoll(p, e, b) _strtoi64(p, e, b)
139 #endif /* _MSC_VER < 1800 */
140 #define strcasecmp(s1, s2) _stricmp(s1, s2)
146 * signbit() is a macro on Linux. Not available on Windows.
149 #define signbit(x) ((x) < 0.0f)
153 /** single-precision inverse square root */
157 /* XXX we could try Quake's fast inverse square root function here */
158 return 1.0F / sqrtf(x);
163 *** LOG2: Log base 2 of float
165 static inline GLfloat LOG2(GLfloat x)
168 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
169 * Based on code from http://www.stereopsis.com/log2.html
171 const GLfloat y = x * x * x * x;
172 const GLuint ix = *((GLuint *) &y);
173 const GLuint exp = (ix >> 23) & 0xFF;
174 const GLint log2 = ((GLint) exp) - 127;
175 return (GLfloat) log2 * (1.0 / 4.0); /* 4, because of x^4 above */
177 /* Pretty fast, and accurate.
178 * Based on code from http://www.flipcode.com/totd/
183 log_2 = ((num.i >> 23) & 255) - 128;
184 num.i &= ~(255 << 23);
186 num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3;
187 return num.f + log_2;
193 *** IS_INF_OR_NAN: test if float is infinite or NaN
195 #if defined(isfinite)
196 #define IS_INF_OR_NAN(x) (!isfinite(x))
197 #elif defined(finite)
198 #define IS_INF_OR_NAN(x) (!finite(x))
199 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
200 #define IS_INF_OR_NAN(x) (!isfinite(x))
202 #define IS_INF_OR_NAN(x) (!finite(x))
207 *** CEILF: ceiling of float
208 *** FLOORF: floor of float
209 *** FABSF: absolute value of float
211 #if defined(__gnu_linux__)
213 #define CEILF(x) ceilf(x)
214 #define FLOORF(x) floorf(x)
215 #define FABSF(x) fabsf(x)
217 #define CEILF(x) ((GLfloat) ceil(x))
218 #define FLOORF(x) ((GLfloat) floor(x))
219 #define FABSF(x) ((GLfloat) fabs(x))
224 * Convert float to int by rounding to nearest integer, away from zero.
226 static inline int IROUND(float f)
228 return (int) ((f >= 0.0F) ? (f + 0.5F) : (f - 0.5F));
233 * Convert float to int64 by rounding to nearest integer.
235 static inline GLint64 IROUND64(float f)
237 return (GLint64) ((f >= 0.0F) ? (f + 0.5F) : (f - 0.5F));
242 * Convert positive float to int by rounding to nearest integer.
244 static inline int IROUND_POS(float f)
247 return (int) (f + 0.5F);
251 # include <xmmintrin.h>
255 * Convert float to int using a fast method. The rounding mode may vary.
257 static inline int F_TO_I(float f)
259 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
261 __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st");
263 #elif defined(USE_X86_ASM) && defined(_MSC_VER)
270 #elif defined(__x86_64__)
271 return _mm_cvt_ss2si(_mm_load_ss(&f));
278 /** Return (as an integer) floor of float */
279 static inline int IFLOOR(float f)
281 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
283 * IEEE floor for computers that round to nearest or even.
284 * 'f' must be between -4194304 and 4194303.
285 * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
286 * but uses some IEEE specific tricks for better speed.
287 * Contributed by Josh Vanderhoof
291 af = (3 << 22) + 0.5 + (double)f;
292 bf = (3 << 22) + 0.5 - (double)f;
293 /* GCC generates an extra fstp/fld without this. */
294 __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
295 __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
296 return (ai - bi) >> 1;
301 af = (3 << 22) + 0.5 + (double)f;
302 bf = (3 << 22) + 0.5 - (double)f;
303 u.f = (float) af; ai = u.i;
304 u.f = (float) bf; bi = u.i;
305 return (ai - bi) >> 1;
310 /** Return (as an integer) ceiling of float */
311 static inline int ICEIL(float f)
313 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
315 * IEEE ceil for computers that round to nearest or even.
316 * 'f' must be between -4194304 and 4194303.
317 * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
318 * but uses some IEEE specific tricks for better speed.
319 * Contributed by Josh Vanderhoof
323 af = (3 << 22) + 0.5 + (double)f;
324 bf = (3 << 22) + 0.5 - (double)f;
325 /* GCC generates an extra fstp/fld without this. */
326 __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
327 __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
328 return (ai - bi + 1) >> 1;
333 af = (3 << 22) + 0.5 + (double)f;
334 bf = (3 << 22) + 0.5 - (double)f;
335 u.f = (float) af; ai = u.i;
336 u.f = (float) bf; bi = u.i;
337 return (ai - bi + 1) >> 1;
343 * Is x a power of two?
346 _mesa_is_pow_two(int x)
348 return !(x & (x - 1));
352 * Round given integer to next higer power of two
353 * If X is zero result is undefined.
355 * Source for the fallback implementation is
356 * Sean Eron Anderson's webpage "Bit Twiddling Hacks"
357 * http://graphics.stanford.edu/~seander/bithacks.html
359 * When using builtin function have to do some work
360 * for case when passed values 1 to prevent hiting
361 * undefined result from __builtin_clz. Undefined
362 * results would be different depending on optimization
363 * level used for build.
365 static inline int32_t
366 _mesa_next_pow_two_32(uint32_t x)
368 #ifdef HAVE___BUILTIN_CLZ
369 uint32_t y = (x != 1);
370 return (1 + y) << ((__builtin_clz(x - y) ^ 31) );
383 static inline int64_t
384 _mesa_next_pow_two_64(uint64_t x)
386 #ifdef HAVE___BUILTIN_CLZLL
387 uint64_t y = (x != 1);
388 STATIC_ASSERT(sizeof(x) == sizeof(long long));
389 return (1 + y) << ((__builtin_clzll(x - y) ^ 63));
405 * Returns the floor form of binary logarithm for a 32-bit integer.
408 _mesa_logbase2(GLuint n)
410 #ifdef HAVE___BUILTIN_CLZ
411 return (31 - __builtin_clz(n | 1));
414 if (n >= 1<<16) { n >>= 16; pos += 16; }
415 if (n >= 1<< 8) { n >>= 8; pos += 8; }
416 if (n >= 1<< 4) { n >>= 4; pos += 4; }
417 if (n >= 1<< 2) { n >>= 2; pos += 2; }
418 if (n >= 1<< 1) { pos += 1; }
425 * Return 1 if this is a little endian machine, 0 if big endian.
427 static inline GLboolean
428 _mesa_little_endian(void)
430 const GLuint ui = 1; /* intentionally not static */
431 return *((const GLubyte *) &ui);
436 /**********************************************************************
441 _mesa_align_malloc( size_t bytes, unsigned long alignment );
444 _mesa_align_calloc( size_t bytes, unsigned long alignment );
447 _mesa_align_free( void *ptr );
450 _mesa_align_realloc(void *oldBuffer, size_t oldSize, size_t newSize,
451 unsigned long alignment);
454 _mesa_exec_malloc( GLuint size );
457 _mesa_exec_free( void *addr );
461 #define FFS_DEFINED 1
462 #ifdef HAVE___BUILTIN_FFS
463 #define ffs __builtin_ffs
465 extern int ffs(int i);
468 #ifdef HAVE___BUILTIN_FFSLL
469 #define ffsll __builtin_ffsll
471 extern int ffsll(long long int i);
473 #endif /* FFS_DEFINED */
476 #ifdef HAVE___BUILTIN_POPCOUNT
477 #define _mesa_bitcount(i) __builtin_popcount(i)
480 _mesa_bitcount(unsigned int n);
483 #ifdef HAVE___BUILTIN_POPCOUNTLL
484 #define _mesa_bitcount_64(i) __builtin_popcountll(i)
487 _mesa_bitcount_64(uint64_t n);
491 * Find the last (most significant) bit set in a word.
493 * Essentially ffs() in the reverse direction.
495 static inline unsigned int
496 _mesa_fls(unsigned int n)
498 #ifdef HAVE___BUILTIN_CLZ
499 return n == 0 ? 0 : 32 - __builtin_clz(n);
514 _mesa_round_to_even(float val);
517 _mesa_float_to_half(float f);
520 _mesa_half_to_float(GLhalfARB h);
523 _mesa_half_is_negative(GLhalfARB h)
529 _mesa_strdup( const char *s );
532 _mesa_str_checksum(const char *str);
535 _mesa_snprintf( char *str, size_t size, const char *fmt, ... ) PRINTFLIKE(3, 4);
538 _mesa_vsnprintf(char *str, size_t size, const char *fmt, va_list arg);
541 #if defined(_MSC_VER) && !defined(snprintf)
542 #define snprintf _snprintf
551 #endif /* IMPORTS_H */