2 * Mesa 3-D graphics library
5 * Copyright (C) 1999-2008 Brian Paul All Rights Reserved.
7 * Permission is hereby granted, free of charge, to any person obtaining a
8 * copy of this software and associated documentation files (the "Software"),
9 * to deal in the Software without restriction, including without limitation
10 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
11 * and/or sell copies of the Software, and to permit persons to whom the
12 * Software is furnished to do so, subject to the following conditions:
14 * The above copyright notice and this permission notice shall be included
15 * in all copies or substantial portions of the Software.
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
18 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20 * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
21 * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
22 * CONNECTION WITH THE SOFTWARE OR THE USE OR 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
; } 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 *** SQRTF: single-precision square root
102 #if 0 /* _mesa_sqrtf() not accurate enough - temporarily disabled */
103 # define SQRTF(X) _mesa_sqrtf(X)
105 # define SQRTF(X) (float) sqrt((float) (X))
110 *** INV_SQRTF: single-precision inverse square root
113 #define INV_SQRTF(X) _mesa_inv_sqrt(X)
115 #define INV_SQRTF(X) (1.0F / SQRTF(X)) /* this is faster on a P4 */
120 * \name Work-arounds for platforms that lack C99 math functions
123 #if (!defined(_XOPEN_SOURCE) || (_XOPEN_SOURCE < 600)) && !defined(_ISOC99_SOURCE) \
124 && (!defined(__STDC_VERSION__) || (__STDC_VERSION__ < 199901L)) \
125 && (!defined(_MSC_VER) || (_MSC_VER < 1400))
126 #define acosf(f) ((float) acos(f))
127 #define asinf(f) ((float) asin(f))
128 #define atan2f(x,y) ((float) atan2(x,y))
129 #define atanf(f) ((float) atan(f))
130 #define cielf(f) ((float) ciel(f))
131 #define cosf(f) ((float) cos(f))
132 #define coshf(f) ((float) cosh(f))
133 #define expf(f) ((float) exp(f))
134 #define exp2f(f) ((float) exp2(f))
135 #define floorf(f) ((float) floor(f))
136 #define logf(f) ((float) log(f))
139 #define log2f(f) (logf(f) * (float) (1.0 / M_LN2))
141 #define log2f(f) ((float) log2(f))
144 #define powf(x,y) ((float) pow(x,y))
145 #define sinf(f) ((float) sin(f))
146 #define sinhf(f) ((float) sinh(f))
147 #define sqrtf(f) ((float) sqrt(f))
148 #define tanf(f) ((float) tan(f))
149 #define tanhf(f) ((float) tanh(f))
150 #define acoshf(f) ((float) acosh(f))
151 #define asinhf(f) ((float) asinh(f))
152 #define atanhf(f) ((float) atanh(f))
155 #if defined(_MSC_VER)
156 static inline float truncf(float x
) { return x
< 0.0f
? ceilf(x
) : floorf(x
); }
157 static inline float exp2f(float x
) { return powf(2.0f
, x
); }
158 static inline float log2f(float x
) { return logf(x
) * 1.442695041f
; }
159 static inline float asinhf(float x
) { return logf(x
+ sqrtf(x
* x
+ 1.0f
)); }
160 static inline float acoshf(float x
) { return logf(x
+ sqrtf(x
* x
- 1.0f
)); }
161 static inline float atanhf(float x
) { return (logf(1.0f
+ x
) - logf(1.0f
- x
)) / 2.0f
; }
162 static inline int isblank(int ch
) { return ch
== ' ' || ch
== '\t'; }
163 #define strtoll(p, e, b) _strtoi64(p, e, b)
168 *** LOG2: Log base 2 of float
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 static inline GLfloat
LOG2(GLfloat x
)
177 const GLfloat y
= x
* x
* x
* x
;
178 const GLuint ix
= *((GLuint
*) &y
);
179 const GLuint exp
= (ix
>> 23) & 0xFF;
180 const GLint log2
= ((GLint
) exp
) - 127;
181 return (GLfloat
) log2
* (1.0 / 4.0); /* 4, because of x^4 above */
184 /* Pretty fast, and accurate.
185 * Based on code from http://www.flipcode.com/totd/
187 static inline GLfloat
LOG2(GLfloat val
)
192 log_2
= ((num
.i
>> 23) & 255) - 128;
193 num
.i
&= ~(255 << 23);
195 num
.f
= ((-1.0f
/3) * num
.f
+ 2) * num
.f
- 2.0f
/3;
196 return num
.f
+ log_2
;
200 * NOTE: log_base_2(x) = log(x) / log(2)
201 * NOTE: 1.442695 = 1/log(2).
203 #define LOG2(x) ((GLfloat) (log(x) * 1.442695F))
208 *** IS_INF_OR_NAN: test if float is infinite or NaN
211 static inline int IS_INF_OR_NAN( float x
)
215 return !(int)((unsigned int)((tmp
.i
& 0x7fffffff)-0x7f800000) >> 31);
217 #elif defined(isfinite)
218 #define IS_INF_OR_NAN(x) (!isfinite(x))
219 #elif defined(finite)
220 #define IS_INF_OR_NAN(x) (!finite(x))
222 #define IS_INF_OR_NAN(x) (!finite(x))
223 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
224 #define IS_INF_OR_NAN(x) (!isfinite(x))
226 #define IS_INF_OR_NAN(x) (!finite(x))
231 *** IS_NEGATIVE: test if float is negative
233 #if defined(USE_IEEE)
234 static inline int GET_FLOAT_BITS( float x
)
240 #define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) < 0)
242 #define IS_NEGATIVE(x) (x < 0.0F)
247 *** DIFFERENT_SIGNS: test if two floats have opposite signs
249 #if defined(USE_IEEE)
250 #define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
252 /* Could just use (x*y<0) except for the flatshading requirements.
253 * Maybe there's a better way?
255 #define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
260 *** CEILF: ceiling of float
261 *** FLOORF: floor of float
262 *** FABSF: absolute value of float
263 *** LOGF: the natural logarithm (base e) of the value
264 *** EXPF: raise e to the value
265 *** LDEXPF: multiply value by an integral power of two
266 *** FREXPF: extract mantissa and exponent from value
268 #if defined(__gnu_linux__)
270 #define CEILF(x) ceilf(x)
271 #define FLOORF(x) floorf(x)
272 #define FABSF(x) fabsf(x)
273 #define LOGF(x) logf(x)
274 #define EXPF(x) expf(x)
275 #define LDEXPF(x,y) ldexpf(x,y)
276 #define FREXPF(x,y) frexpf(x,y)
278 #define CEILF(x) ((GLfloat) ceil(x))
279 #define FLOORF(x) ((GLfloat) floor(x))
280 #define FABSF(x) ((GLfloat) fabs(x))
281 #define LOGF(x) ((GLfloat) log(x))
282 #define EXPF(x) ((GLfloat) exp(x))
283 #define LDEXPF(x,y) ((GLfloat) ldexp(x,y))
284 #define FREXPF(x,y) ((GLfloat) frexp(x,y))
289 * Convert float to int by rounding to nearest integer, away from zero.
291 static inline int IROUND(float f
)
293 return (int) ((f
>= 0.0F
) ? (f
+ 0.5F
) : (f
- 0.5F
));
298 * Convert float to int64 by rounding to nearest integer.
300 static inline GLint64
IROUND64(float f
)
302 return (GLint64
) ((f
>= 0.0F
) ? (f
+ 0.5F
) : (f
- 0.5F
));
307 * Convert positive float to int by rounding to nearest integer.
309 static inline int IROUND_POS(float f
)
312 return (int) (f
+ 0.5F
);
317 * Convert float to int using a fast method. The rounding mode may vary.
318 * XXX We could use an x86-64/SSE2 version here.
320 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
321 static inline int F_TO_I(float f
)
324 __asm__ ("fistpl %0" : "=m" (r
) : "t" (f
) : "st");
327 #elif defined(USE_X86_ASM) && defined(_MSC_VER)
328 static inline int F_TO_I(float f
)
337 #elif defined(__WATCOMC__) && defined(__386__)
338 long F_TO_I(float f
);
339 #pragma aux iround = \
341 "fistp dword ptr [esp]" \
347 #define F_TO_I(f) IROUND(f)
352 *** IFLOOR: return (as an integer) floor of float
354 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
356 * IEEE floor for computers that round to nearest or even.
357 * 'f' must be between -4194304 and 4194303.
358 * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
359 * but uses some IEEE specific tricks for better speed.
360 * Contributed by Josh Vanderhoof
362 static inline int ifloor(float f
)
366 af
= (3 << 22) + 0.5 + (double)f
;
367 bf
= (3 << 22) + 0.5 - (double)f
;
368 /* GCC generates an extra fstp/fld without this. */
369 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
370 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
371 return (ai
- bi
) >> 1;
373 #define IFLOOR(x) ifloor(x)
374 #elif defined(USE_IEEE)
375 static inline int ifloor(float f
)
381 af
= (3 << 22) + 0.5 + (double)f
;
382 bf
= (3 << 22) + 0.5 - (double)f
;
383 u
.f
= (float) af
; ai
= u
.i
;
384 u
.f
= (float) bf
; bi
= u
.i
;
385 return (ai
- bi
) >> 1;
387 #define IFLOOR(x) ifloor(x)
389 static inline int ifloor(float f
)
392 return (i
> f
) ? i
- 1 : i
;
394 #define IFLOOR(x) ifloor(x)
399 *** ICEIL: return (as an integer) ceiling of float
401 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
403 * IEEE ceil for computers that round to nearest or even.
404 * 'f' must be between -4194304 and 4194303.
405 * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
406 * but uses some IEEE specific tricks for better speed.
407 * Contributed by Josh Vanderhoof
409 static inline int iceil(float f
)
413 af
= (3 << 22) + 0.5 + (double)f
;
414 bf
= (3 << 22) + 0.5 - (double)f
;
415 /* GCC generates an extra fstp/fld without this. */
416 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
417 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
418 return (ai
- bi
+ 1) >> 1;
420 #define ICEIL(x) iceil(x)
421 #elif defined(USE_IEEE)
422 static inline int iceil(float f
)
427 af
= (3 << 22) + 0.5 + (double)f
;
428 bf
= (3 << 22) + 0.5 - (double)f
;
429 u
.f
= (float) af
; ai
= u
.i
;
430 u
.f
= (float) bf
; bi
= u
.i
;
431 return (ai
- bi
+ 1) >> 1;
433 #define ICEIL(x) iceil(x)
435 static inline int iceil(float f
)
438 return (i
< f
) ? i
+ 1 : i
;
440 #define ICEIL(x) iceil(x)
445 * Is x a power of two?
448 _mesa_is_pow_two(int x
)
450 return !(x
& (x
- 1));
454 * Round given integer to next higer power of two
455 * If X is zero result is undefined.
457 * Source for the fallback implementation is
458 * Sean Eron Anderson's webpage "Bit Twiddling Hacks"
459 * http://graphics.stanford.edu/~seander/bithacks.html
461 * When using builtin function have to do some work
462 * for case when passed values 1 to prevent hiting
463 * undefined result from __builtin_clz. Undefined
464 * results would be different depending on optimization
465 * level used for build.
467 static inline int32_t
468 _mesa_next_pow_two_32(uint32_t x
)
470 #if defined(__GNUC__) && \
471 ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
472 uint32_t y
= (x
!= 1);
473 return (1 + y
) << ((__builtin_clz(x
- y
) ^ 31) );
486 static inline int64_t
487 _mesa_next_pow_two_64(uint64_t x
)
489 #if defined(__GNUC__) && \
490 ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
491 uint64_t y
= (x
!= 1);
492 if (sizeof(x
) == sizeof(long))
493 return (1 + y
) << ((__builtin_clzl(x
- y
) ^ 63));
495 return (1 + y
) << ((__builtin_clzll(x
- y
) ^ 63));
511 * Returns the floor form of binary logarithm for a 32-bit integer.
514 _mesa_logbase2(GLuint n
)
516 #if defined(__GNUC__) && \
517 ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
518 return (31 - __builtin_clz(n
| 1));
521 if (n
>= 1<<16) { n
>>= 16; pos
+= 16; }
522 if (n
>= 1<< 8) { n
>>= 8; pos
+= 8; }
523 if (n
>= 1<< 4) { n
>>= 4; pos
+= 4; }
524 if (n
>= 1<< 2) { n
>>= 2; pos
+= 2; }
525 if (n
>= 1<< 1) { pos
+= 1; }
532 * Return 1 if this is a little endian machine, 0 if big endian.
534 static inline GLboolean
535 _mesa_little_endian(void)
537 const GLuint ui
= 1; /* intentionally not static */
538 return *((const GLubyte
*) &ui
);
543 /**********************************************************************
548 _mesa_align_malloc( size_t bytes
, unsigned long alignment
);
551 _mesa_align_calloc( size_t bytes
, unsigned long alignment
);
554 _mesa_align_free( void *ptr
);
557 _mesa_align_realloc(void *oldBuffer
, size_t oldSize
, size_t newSize
,
558 unsigned long alignment
);
561 _mesa_exec_malloc( GLuint size
);
564 _mesa_exec_free( void *addr
);
567 _mesa_realloc( void *oldBuffer
, size_t oldSize
, size_t newSize
);
570 _mesa_memset16( unsigned short *dst
, unsigned short val
, size_t n
);
573 _mesa_sqrtd(double x
);
576 _mesa_sqrtf(float x
);
579 _mesa_inv_sqrtf(float x
);
582 _mesa_init_sqrt_table(void);
586 #define FFS_DEFINED 1
589 #if defined(__MINGW32__) || defined(__CYGWIN__) || defined(ANDROID) || defined(__APPLE__)
590 #define ffs __builtin_ffs
591 #define ffsll __builtin_ffsll
596 extern int ffs(int i
);
597 extern int ffsll(long long int i
);
599 #endif /*__ GNUC__ */
600 #endif /* FFS_DEFINED */
603 #if defined(__GNUC__) && ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
604 #define _mesa_bitcount(i) __builtin_popcount(i)
605 #define _mesa_bitcount_64(i) __builtin_popcountll(i)
608 _mesa_bitcount(unsigned int n
);
610 _mesa_bitcount_64(uint64_t n
);
615 _mesa_float_to_half(float f
);
618 _mesa_half_to_float(GLhalfARB h
);
622 _mesa_bsearch( const void *key
, const void *base
, size_t nmemb
, size_t size
,
623 int (*compar
)(const void *, const void *) );
626 _mesa_getenv( const char *var
);
629 _mesa_strdup( const char *s
);
632 _mesa_strtof( const char *s
, char **end
);
635 _mesa_str_checksum(const char *str
);
638 _mesa_snprintf( char *str
, size_t size
, const char *fmt
, ... ) PRINTFLIKE(3, 4);
641 _mesa_vsnprintf(char *str
, size_t size
, const char *fmt
, va_list arg
);
644 #if defined(_MSC_VER) && !defined(snprintf)
645 #define snprintf _snprintf
650 * On Mingw32 we need to use __mingw_fprintf() to parse formats such
651 * as "0x%llx", and possibly others
654 #define fprintf __mingw_fprintf
664 #endif /* IMPORTS_H */