53e40b44546278366d9824cb1d2561a5f5e825cc
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 \p BYTES bytes */
53 #define MALLOC(BYTES) malloc(BYTES)
54 /** Allocate and zero \p BYTES bytes */
55 #define CALLOC(BYTES) calloc(1, BYTES)
56 /** Allocate a structure of type \p T */
57 #define MALLOC_STRUCT(T) (struct T *) malloc(sizeof(struct T))
58 /** Allocate and zero a structure of type \p T */
59 #define CALLOC_STRUCT(T) (struct T *) calloc(1, sizeof(struct T))
61 #define FREE(PTR) free(PTR)
67 * For GL_ARB_vertex_buffer_object we need to treat vertex array pointers
68 * as offsets into buffer stores. Since the vertex array pointer and
69 * buffer store pointer are both pointers and we need to add them, we use
71 * Both pointers/offsets are expressed in bytes.
73 #define ADD_POINTERS(A, B) ( (GLubyte *) (A) + (uintptr_t) (B) )
77 * Sometimes we treat GLfloats as GLints. On x86 systems, moving a float
78 * as a int (thereby using integer registers instead of FP registers) is
79 * a performance win. Typically, this can be done with ordinary casts.
80 * But with gcc's -fstrict-aliasing flag (which defaults to on in gcc 3.0)
81 * these casts generate warnings.
82 * The following union typedef is used to solve that.
84 typedef union { GLfloat f
; GLint i
; GLuint u
; } fi_type
;
88 /**********************************************************************
92 #define MAX_GLUSHORT 0xffff
93 #define MAX_GLUINT 0xffffffff
95 /* Degrees to radians conversion: */
96 #define DEG2RAD (M_PI/180.0)
100 * \name Work-arounds for platforms that lack C99 math functions
103 #if (!defined(_XOPEN_SOURCE) || (_XOPEN_SOURCE < 600)) && !defined(_ISOC99_SOURCE) \
104 && (!defined(__STDC_VERSION__) || (__STDC_VERSION__ < 199901L)) \
105 && (!defined(_MSC_VER) || (_MSC_VER < 1400))
106 #define acosf(f) ((float) acos(f))
107 #define asinf(f) ((float) asin(f))
108 #define atan2f(x,y) ((float) atan2(x,y))
109 #define atanf(f) ((float) atan(f))
110 #define ceilf(f) ((float) ceil(f))
111 #define cosf(f) ((float) cos(f))
112 #define coshf(f) ((float) cosh(f))
113 #define expf(f) ((float) exp(f))
114 #define exp2f(f) ((float) exp2(f))
115 #define floorf(f) ((float) floor(f))
116 #define logf(f) ((float) log(f))
119 #define log2f(f) (logf(f) * (float) (1.0 / M_LN2))
121 #define log2f(f) ((float) log2(f))
124 #define powf(x,y) ((float) pow(x,y))
125 #define sinf(f) ((float) sin(f))
126 #define sinhf(f) ((float) sinh(f))
127 #define sqrtf(f) ((float) sqrt(f))
128 #define tanf(f) ((float) tan(f))
129 #define tanhf(f) ((float) tanh(f))
130 #define acoshf(f) ((float) acosh(f))
131 #define asinhf(f) ((float) asinh(f))
132 #define atanhf(f) ((float) atanh(f))
135 #if defined(_MSC_VER)
136 static inline float truncf(float x
) { return x
< 0.0f
? ceilf(x
) : floorf(x
); }
137 static inline float exp2f(float x
) { return powf(2.0f
, x
); }
138 static inline float log2f(float x
) { return logf(x
) * 1.442695041f
; }
139 static inline float asinhf(float x
) { return logf(x
+ sqrtf(x
* x
+ 1.0f
)); }
140 static inline float acoshf(float x
) { return logf(x
+ sqrtf(x
* x
- 1.0f
)); }
141 static inline float atanhf(float x
) { return (logf(1.0f
+ x
) - logf(1.0f
- x
)) / 2.0f
; }
142 static inline int isblank(int ch
) { return ch
== ' ' || ch
== '\t'; }
143 #define strtoll(p, e, b) _strtoi64(p, e, b)
149 * signbit() is a macro on Linux. Not available on Windows.
152 #define signbit(x) ((x) < 0.0f)
156 /** single-precision inverse square root */
160 /* XXX we could try Quake's fast inverse square root function here */
161 return 1.0F
/ sqrtf(x
);
166 *** LOG2: Log base 2 of float
168 static inline GLfloat
LOG2(GLfloat x
)
172 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
173 * Based on code from http://www.stereopsis.com/log2.html
175 const GLfloat y
= x
* x
* x
* x
;
176 const GLuint ix
= *((GLuint
*) &y
);
177 const GLuint exp
= (ix
>> 23) & 0xFF;
178 const GLint log2
= ((GLint
) exp
) - 127;
179 return (GLfloat
) log2
* (1.0 / 4.0); /* 4, because of x^4 above */
181 /* Pretty fast, and accurate.
182 * Based on code from http://www.flipcode.com/totd/
187 log_2
= ((num
.i
>> 23) & 255) - 128;
188 num
.i
&= ~(255 << 23);
190 num
.f
= ((-1.0f
/3) * num
.f
+ 2) * num
.f
- 2.0f
/3;
191 return num
.f
+ log_2
;
194 * NOTE: log_base_2(x) = log(x) / log(2)
195 * NOTE: 1.442695 = 1/log(2).
197 return (GLfloat
) (log(x
) * 1.442695F
);
204 *** IS_INF_OR_NAN: test if float is infinite or NaN
207 static inline int IS_INF_OR_NAN( float x
)
211 return !(int)((unsigned int)((tmp
.i
& 0x7fffffff)-0x7f800000) >> 31);
213 #elif defined(isfinite)
214 #define IS_INF_OR_NAN(x) (!isfinite(x))
215 #elif defined(finite)
216 #define IS_INF_OR_NAN(x) (!finite(x))
217 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
218 #define IS_INF_OR_NAN(x) (!isfinite(x))
220 #define IS_INF_OR_NAN(x) (!finite(x))
225 *** CEILF: ceiling of float
226 *** FLOORF: floor of float
227 *** FABSF: absolute value of float
228 *** LOGF: the natural logarithm (base e) of the value
229 *** EXPF: raise e to the value
230 *** LDEXPF: multiply value by an integral power of two
231 *** FREXPF: extract mantissa and exponent from value
233 #if defined(__gnu_linux__)
235 #define CEILF(x) ceilf(x)
236 #define FLOORF(x) floorf(x)
237 #define FABSF(x) fabsf(x)
238 #define LOGF(x) logf(x)
239 #define EXPF(x) expf(x)
240 #define LDEXPF(x,y) ldexpf(x,y)
241 #define FREXPF(x,y) frexpf(x,y)
243 #define CEILF(x) ((GLfloat) ceil(x))
244 #define FLOORF(x) ((GLfloat) floor(x))
245 #define FABSF(x) ((GLfloat) fabs(x))
246 #define LOGF(x) ((GLfloat) log(x))
247 #define EXPF(x) ((GLfloat) exp(x))
248 #define LDEXPF(x,y) ((GLfloat) ldexp(x,y))
249 #define FREXPF(x,y) ((GLfloat) frexp(x,y))
254 * Convert float to int by rounding to nearest integer, away from zero.
256 static inline int IROUND(float f
)
258 return (int) ((f
>= 0.0F
) ? (f
+ 0.5F
) : (f
- 0.5F
));
263 * Convert float to int64 by rounding to nearest integer.
265 static inline GLint64
IROUND64(float f
)
267 return (GLint64
) ((f
>= 0.0F
) ? (f
+ 0.5F
) : (f
- 0.5F
));
272 * Convert positive float to int by rounding to nearest integer.
274 static inline int IROUND_POS(float f
)
277 return (int) (f
+ 0.5F
);
282 * Convert float to int using a fast method. The rounding mode may vary.
283 * XXX We could use an x86-64/SSE2 version here.
285 static inline int F_TO_I(float f
)
287 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
289 __asm__ ("fistpl %0" : "=m" (r
) : "t" (f
) : "st");
291 #elif defined(USE_X86_ASM) && defined(_MSC_VER)
304 /** Return (as an integer) floor of float */
305 static inline int IFLOOR(float f
)
307 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
309 * IEEE floor for computers that round to nearest or even.
310 * 'f' must be between -4194304 and 4194303.
311 * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
312 * but uses some IEEE specific tricks for better speed.
313 * Contributed by Josh Vanderhoof
317 af
= (3 << 22) + 0.5 + (double)f
;
318 bf
= (3 << 22) + 0.5 - (double)f
;
319 /* GCC generates an extra fstp/fld without this. */
320 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
321 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
322 return (ai
- bi
) >> 1;
323 #elif defined(USE_IEEE)
327 af
= (3 << 22) + 0.5 + (double)f
;
328 bf
= (3 << 22) + 0.5 - (double)f
;
329 u
.f
= (float) af
; ai
= u
.i
;
330 u
.f
= (float) bf
; bi
= u
.i
;
331 return (ai
- bi
) >> 1;
334 return (i
> f
) ? i
- 1 : i
;
339 /** Return (as an integer) ceiling of float */
340 static inline int ICEIL(float f
)
342 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
344 * IEEE ceil for computers that round to nearest or even.
345 * 'f' must be between -4194304 and 4194303.
346 * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
347 * but uses some IEEE specific tricks for better speed.
348 * Contributed by Josh Vanderhoof
352 af
= (3 << 22) + 0.5 + (double)f
;
353 bf
= (3 << 22) + 0.5 - (double)f
;
354 /* GCC generates an extra fstp/fld without this. */
355 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
356 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
357 return (ai
- bi
+ 1) >> 1;
358 #elif defined(USE_IEEE)
362 af
= (3 << 22) + 0.5 + (double)f
;
363 bf
= (3 << 22) + 0.5 - (double)f
;
364 u
.f
= (float) af
; ai
= u
.i
;
365 u
.f
= (float) bf
; bi
= u
.i
;
366 return (ai
- bi
+ 1) >> 1;
369 return (i
< f
) ? i
+ 1 : i
;
375 * Is x a power of two?
378 _mesa_is_pow_two(int x
)
380 return !(x
& (x
- 1));
384 * Round given integer to next higer power of two
385 * If X is zero result is undefined.
387 * Source for the fallback implementation is
388 * Sean Eron Anderson's webpage "Bit Twiddling Hacks"
389 * http://graphics.stanford.edu/~seander/bithacks.html
391 * When using builtin function have to do some work
392 * for case when passed values 1 to prevent hiting
393 * undefined result from __builtin_clz. Undefined
394 * results would be different depending on optimization
395 * level used for build.
397 static inline int32_t
398 _mesa_next_pow_two_32(uint32_t x
)
400 #if defined(__GNUC__) && \
401 ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
402 uint32_t y
= (x
!= 1);
403 return (1 + y
) << ((__builtin_clz(x
- y
) ^ 31) );
416 static inline int64_t
417 _mesa_next_pow_two_64(uint64_t x
)
419 #if defined(__GNUC__) && \
420 ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
421 uint64_t y
= (x
!= 1);
422 if (sizeof(x
) == sizeof(long))
423 return (1 + y
) << ((__builtin_clzl(x
- y
) ^ 63));
425 return (1 + y
) << ((__builtin_clzll(x
- y
) ^ 63));
441 * Returns the floor form of binary logarithm for a 32-bit integer.
444 _mesa_logbase2(GLuint n
)
446 #if defined(__GNUC__) && \
447 ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
448 return (31 - __builtin_clz(n
| 1));
451 if (n
>= 1<<16) { n
>>= 16; pos
+= 16; }
452 if (n
>= 1<< 8) { n
>>= 8; pos
+= 8; }
453 if (n
>= 1<< 4) { n
>>= 4; pos
+= 4; }
454 if (n
>= 1<< 2) { n
>>= 2; pos
+= 2; }
455 if (n
>= 1<< 1) { pos
+= 1; }
462 * Return 1 if this is a little endian machine, 0 if big endian.
464 static inline GLboolean
465 _mesa_little_endian(void)
467 const GLuint ui
= 1; /* intentionally not static */
468 return *((const GLubyte
*) &ui
);
473 /**********************************************************************
478 _mesa_align_malloc( size_t bytes
, unsigned long alignment
);
481 _mesa_align_calloc( size_t bytes
, unsigned long alignment
);
484 _mesa_align_free( void *ptr
);
487 _mesa_align_realloc(void *oldBuffer
, size_t oldSize
, size_t newSize
,
488 unsigned long alignment
);
491 _mesa_exec_malloc( GLuint size
);
494 _mesa_exec_free( void *addr
);
497 _mesa_realloc( void *oldBuffer
, size_t oldSize
, size_t newSize
);
501 #define FFS_DEFINED 1
503 #define ffs __builtin_ffs
504 #define ffsll __builtin_ffsll
506 extern int ffs(int i
);
507 extern int ffsll(long long int i
);
508 #endif /*__ GNUC__ */
509 #endif /* FFS_DEFINED */
512 #if defined(__GNUC__) && ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
513 #define _mesa_bitcount(i) __builtin_popcount(i)
514 #define _mesa_bitcount_64(i) __builtin_popcountll(i)
517 _mesa_bitcount(unsigned int n
);
519 _mesa_bitcount_64(uint64_t n
);
523 * Find the last (most significant) bit set in a word.
525 * Essentially ffs() in the reverse direction.
527 static inline unsigned int
528 _mesa_fls(unsigned int n
)
530 #if defined(__GNUC__) && ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304)
531 return n
== 0 ? 0 : 32 - __builtin_clz(n
);
546 _mesa_round_to_even(float val
);
549 _mesa_float_to_half(float f
);
552 _mesa_half_to_float(GLhalfARB h
);
556 _mesa_bsearch( const void *key
, const void *base
, size_t nmemb
, size_t size
,
557 int (*compar
)(const void *, const void *) );
560 _mesa_getenv( const char *var
);
563 _mesa_strdup( const char *s
);
566 _mesa_strtof( const char *s
, char **end
);
569 _mesa_str_checksum(const char *str
);
572 _mesa_snprintf( char *str
, size_t size
, const char *fmt
, ... ) PRINTFLIKE(3, 4);
575 _mesa_vsnprintf(char *str
, size_t size
, const char *fmt
, va_list arg
);
578 #if defined(_MSC_VER) && !defined(snprintf)
579 #define snprintf _snprintf
588 #endif /* IMPORTS_H */