c4e28df0513c4a56ad5bbf47e7e22f0a8e88a815
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) _mesa_malloc(BYTES)
54 /** Allocate and zero \p BYTES bytes */
55 #define CALLOC(BYTES) _mesa_calloc(BYTES)
56 /** Allocate a structure of type \p T */
57 #define MALLOC_STRUCT(T) (struct T *) _mesa_malloc(sizeof(struct T))
58 /** Allocate and zero a structure of type \p T */
59 #define CALLOC_STRUCT(T) (struct T *) _mesa_calloc(sizeof(struct T))
61 #define FREE(PTR) _mesa_free(PTR)
63 /** Allocate \p BYTES aligned at \p N bytes */
64 #define ALIGN_MALLOC(BYTES, N) _mesa_align_malloc(BYTES, N)
65 /** Allocate and zero \p BYTES bytes aligned at \p N bytes */
66 #define ALIGN_CALLOC(BYTES, N) _mesa_align_calloc(BYTES, N)
67 /** Allocate a structure of type \p T aligned at \p N bytes */
68 #define ALIGN_MALLOC_STRUCT(T, N) (struct T *) _mesa_align_malloc(sizeof(struct T), N)
69 /** Allocate and zero a structure of type \p T aligned at \p N bytes */
70 #define ALIGN_CALLOC_STRUCT(T, N) (struct T *) _mesa_align_calloc(sizeof(struct T), N)
71 /** Free aligned memory */
72 #define ALIGN_FREE(PTR) _mesa_align_free(PTR)
74 /** Set \p N bytes in \p DST to \p VAL */
75 #define MEMSET( DST, VAL, N ) memset(DST, VAL, N)
81 * For GL_ARB_vertex_buffer_object we need to treat vertex array pointers
82 * as offsets into buffer stores. Since the vertex array pointer and
83 * buffer store pointer are both pointers and we need to add them, we use
85 * Both pointers/offsets are expressed in bytes.
87 #define ADD_POINTERS(A, B) ( (GLubyte *) (A) + (uintptr_t) (B) )
91 * Sometimes we treat GLfloats as GLints. On x86 systems, moving a float
92 * as a int (thereby using integer registers instead of FP registers) is
93 * a performance win. Typically, this can be done with ordinary casts.
94 * But with gcc's -fstrict-aliasing flag (which defaults to on in gcc 3.0)
95 * these casts generate warnings.
96 * The following union typedef is used to solve that.
98 typedef union { GLfloat f
; GLint i
; } fi_type
;
102 /**********************************************************************
106 #define MAX_GLUSHORT 0xffff
107 #define MAX_GLUINT 0xffffffff
109 /* Degrees to radians conversion: */
110 #define DEG2RAD (M_PI/180.0)
114 *** SQRTF: single-precision square root
116 #if 0 /* _mesa_sqrtf() not accurate enough - temporarily disabled */
117 # define SQRTF(X) _mesa_sqrtf(X)
119 # define SQRTF(X) (float) sqrt((float) (X))
124 *** INV_SQRTF: single-precision inverse square root
127 #define INV_SQRTF(X) _mesa_inv_sqrt(X)
129 #define INV_SQRTF(X) (1.0F / SQRTF(X)) /* this is faster on a P4 */
134 *** LOG2: Log base 2 of float
138 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
139 * Based on code from http://www.stereopsis.com/log2.html
141 static INLINE GLfloat
LOG2(GLfloat x
)
143 const GLfloat y
= x
* x
* x
* x
;
144 const GLuint ix
= *((GLuint
*) &y
);
145 const GLuint exp
= (ix
>> 23) & 0xFF;
146 const GLint log2
= ((GLint
) exp
) - 127;
147 return (GLfloat
) log2
* (1.0 / 4.0); /* 4, because of x^4 above */
150 /* Pretty fast, and accurate.
151 * Based on code from http://www.flipcode.com/totd/
153 static INLINE GLfloat
LOG2(GLfloat val
)
158 log_2
= ((num
.i
>> 23) & 255) - 128;
159 num
.i
&= ~(255 << 23);
161 num
.f
= ((-1.0f
/3) * num
.f
+ 2) * num
.f
- 2.0f
/3;
162 return num
.f
+ log_2
;
166 * NOTE: log_base_2(x) = log(x) / log(2)
167 * NOTE: 1.442695 = 1/log(2).
169 #define LOG2(x) ((GLfloat) (log(x) * 1.442695F))
174 *** IS_INF_OR_NAN: test if float is infinite or NaN
177 static INLINE
int IS_INF_OR_NAN( float x
)
181 return !(int)((unsigned int)((tmp
.i
& 0x7fffffff)-0x7f800000) >> 31);
183 #elif defined(isfinite)
184 #define IS_INF_OR_NAN(x) (!isfinite(x))
185 #elif defined(finite)
186 #define IS_INF_OR_NAN(x) (!finite(x))
188 #define IS_INF_OR_NAN(x) (!finite(x))
189 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
190 #define IS_INF_OR_NAN(x) (!isfinite(x))
192 #define IS_INF_OR_NAN(x) (!finite(x))
197 *** IS_NEGATIVE: test if float is negative
199 #if defined(USE_IEEE)
200 static INLINE
int GET_FLOAT_BITS( float x
)
206 #define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) < 0)
208 #define IS_NEGATIVE(x) (x < 0.0F)
213 *** DIFFERENT_SIGNS: test if two floats have opposite signs
215 #if defined(USE_IEEE)
216 #define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
218 /* Could just use (x*y<0) except for the flatshading requirements.
219 * Maybe there's a better way?
221 #define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
226 *** CEILF: ceiling of float
227 *** FLOORF: floor of float
228 *** FABSF: absolute value of float
229 *** LOGF: the natural logarithm (base e) of the value
230 *** EXPF: raise e to the value
231 *** LDEXPF: multiply value by an integral power of two
232 *** FREXPF: extract mantissa and exponent from value
234 #if defined(__gnu_linux__)
236 #define CEILF(x) ceilf(x)
237 #define FLOORF(x) floorf(x)
238 #define FABSF(x) fabsf(x)
239 #define LOGF(x) logf(x)
240 #define EXPF(x) expf(x)
241 #define LDEXPF(x,y) ldexpf(x,y)
242 #define FREXPF(x,y) frexpf(x,y)
244 #define CEILF(x) ((GLfloat) ceil(x))
245 #define FLOORF(x) ((GLfloat) floor(x))
246 #define FABSF(x) ((GLfloat) fabs(x))
247 #define LOGF(x) ((GLfloat) log(x))
248 #define EXPF(x) ((GLfloat) exp(x))
249 #define LDEXPF(x,y) ((GLfloat) ldexp(x,y))
250 #define FREXPF(x,y) ((GLfloat) frexp(x,y))
255 *** IROUND: return (as an integer) float rounded to nearest integer
257 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__) && \
258 (!(defined(__BEOS__) || defined(__HAIKU__)) || \
259 (__GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 95)))
260 static INLINE
int iround(float f
)
263 __asm__ ("fistpl %0" : "=m" (r
) : "t" (f
) : "st");
266 #define IROUND(x) iround(x)
267 #elif defined(USE_X86_ASM) && defined(_MSC_VER)
268 static INLINE
int iround(float f
)
277 #define IROUND(x) iround(x)
278 #elif defined(__WATCOMC__) && defined(__386__)
279 long iround(float f
);
280 #pragma aux iround = \
282 "fistp dword ptr [esp]" \
287 #define IROUND(x) iround(x)
289 #define IROUND(f) ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
292 #define IROUND64(f) ((GLint64) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
295 *** IROUND_POS: return (as an integer) positive float rounded to nearest int
298 #define IROUND_POS(f) (assert((f) >= 0.0F), IROUND(f))
300 #define IROUND_POS(f) (IROUND(f))
305 *** IFLOOR: return (as an integer) floor of float
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
315 static INLINE
int ifloor(float f
)
319 af
= (3 << 22) + 0.5 + (double)f
;
320 bf
= (3 << 22) + 0.5 - (double)f
;
321 /* GCC generates an extra fstp/fld without this. */
322 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
323 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
324 return (ai
- bi
) >> 1;
326 #define IFLOOR(x) ifloor(x)
327 #elif defined(USE_IEEE)
328 static INLINE
int ifloor(float f
)
334 af
= (3 << 22) + 0.5 + (double)f
;
335 bf
= (3 << 22) + 0.5 - (double)f
;
336 u
.f
= (float) af
; ai
= u
.i
;
337 u
.f
= (float) bf
; bi
= u
.i
;
338 return (ai
- bi
) >> 1;
340 #define IFLOOR(x) ifloor(x)
342 static INLINE
int ifloor(float f
)
345 return (i
> f
) ? i
- 1 : i
;
347 #define IFLOOR(x) ifloor(x)
352 *** ICEIL: return (as an integer) ceiling of float
354 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
356 * IEEE ceil for computers that round to nearest or even.
357 * 'f' must be between -4194304 and 4194303.
358 * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
359 * but uses some IEEE specific tricks for better speed.
360 * Contributed by Josh Vanderhoof
362 static INLINE
int iceil(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) >> 1;
373 #define ICEIL(x) iceil(x)
374 #elif defined(USE_IEEE)
375 static INLINE
int iceil(float f
)
380 af
= (3 << 22) + 0.5 + (double)f
;
381 bf
= (3 << 22) + 0.5 - (double)f
;
382 u
.f
= (float) af
; ai
= u
.i
;
383 u
.f
= (float) bf
; bi
= u
.i
;
384 return (ai
- bi
+ 1) >> 1;
386 #define ICEIL(x) iceil(x)
388 static INLINE
int iceil(float f
)
391 return (i
< f
) ? i
+ 1 : i
;
393 #define ICEIL(x) iceil(x)
398 * Is x a power of two?
401 _mesa_is_pow_two(int x
)
403 return !(x
& (x
- 1));
407 * Round given integer to next higer power of two
408 * If X is zero result is undefined.
410 * Source for the fallback implementation is
411 * Sean Eron Anderson's webpage "Bit Twiddling Hacks"
412 * http://graphics.stanford.edu/~seander/bithacks.html
414 * When using builtin function have to do some work
415 * for case when passed values 1 to prevent hiting
416 * undefined result from __builtin_clz. Undefined
417 * results would be different depending on optimization
418 * level used for build.
420 static INLINE
int32_t
421 _mesa_next_pow_two_32(uint32_t x
)
424 uint32_t y
= (x
!= 1);
425 return (1 + y
) << ((__builtin_clz(x
- y
) ^ 31) );
438 static INLINE
int64_t
439 _mesa_next_pow_two_64(uint64_t x
)
442 uint64_t y
= (x
!= 1);
443 if (sizeof(x
) == sizeof(long))
444 return (1 + y
) << ((__builtin_clzl(x
- y
) ^ 63));
446 return (1 + y
) << ((__builtin_clzll(x
- y
) ^ 63));
462 *** UNCLAMPED_FLOAT_TO_UBYTE: clamp float to [0,1] and map to ubyte in [0,255]
463 *** CLAMPED_FLOAT_TO_UBYTE: map float known to be in [0,1] to ubyte in [0,255]
465 #if defined(USE_IEEE) && !defined(DEBUG)
466 #define IEEE_0996 0x3f7f0000 /* 0.996 or so */
467 /* This function/macro is sensitive to precision. Test very carefully
470 #define UNCLAMPED_FLOAT_TO_UBYTE(UB, F) \
476 else if (__tmp.i >= IEEE_0996) \
477 UB = (GLubyte) 255; \
479 __tmp.f = __tmp.f * (255.0F/256.0F) + 32768.0F; \
480 UB = (GLubyte) __tmp.i; \
483 #define CLAMPED_FLOAT_TO_UBYTE(UB, F) \
486 __tmp.f = (F) * (255.0F/256.0F) + 32768.0F; \
487 UB = (GLubyte) __tmp.i; \
490 #define UNCLAMPED_FLOAT_TO_UBYTE(ub, f) \
491 ub = ((GLubyte) IROUND(CLAMP((f), 0.0F, 1.0F) * 255.0F))
492 #define CLAMPED_FLOAT_TO_UBYTE(ub, f) \
493 ub = ((GLubyte) IROUND((f) * 255.0F))
498 * Return 1 if this is a little endian machine, 0 if big endian.
500 static INLINE GLboolean
501 _mesa_little_endian(void)
503 const GLuint ui
= 1; /* intentionally not static */
504 return *((const GLubyte
*) &ui
);
509 /**********************************************************************
514 _mesa_malloc( size_t bytes
);
517 _mesa_calloc( size_t bytes
);
520 _mesa_free( void *ptr
);
523 _mesa_align_malloc( size_t bytes
, unsigned long alignment
);
526 _mesa_align_calloc( size_t bytes
, unsigned long alignment
);
529 _mesa_align_free( void *ptr
);
532 _mesa_align_realloc(void *oldBuffer
, size_t oldSize
, size_t newSize
,
533 unsigned long alignment
);
536 _mesa_exec_malloc( GLuint size
);
539 _mesa_exec_free( void *addr
);
542 _mesa_realloc( void *oldBuffer
, size_t oldSize
, size_t newSize
);
545 _mesa_memset16( unsigned short *dst
, unsigned short val
, size_t n
);
548 _mesa_bzero( void *dst
, size_t n
);
560 _mesa_asinf(float x
);
563 _mesa_atanf(float x
);
566 _mesa_sqrtd(double x
);
569 _mesa_sqrtf(float x
);
572 _mesa_inv_sqrtf(float x
);
575 _mesa_init_sqrt_table(void);
578 _mesa_pow(double x
, double y
);
581 _mesa_ffs(int32_t i
);
584 _mesa_ffsll(int64_t i
);
587 _mesa_bitcount(unsigned int n
);
590 _mesa_float_to_half(float f
);
593 _mesa_half_to_float(GLhalfARB h
);
597 _mesa_bsearch( const void *key
, const void *base
, size_t nmemb
, size_t size
,
598 int (*compar
)(const void *, const void *) );
601 _mesa_getenv( const char *var
);
604 _mesa_strdup( const char *s
);
607 _mesa_strtod( const char *s
, char **end
);
610 _mesa_str_checksum(const char *str
);
613 _mesa_sprintf( char *str
, const char *fmt
, ... );
616 _mesa_snprintf( char *str
, size_t size
, const char *fmt
, ... );
619 _mesa_printf( const char *fmtString
, ... );
622 _mesa_fprintf( FILE *f
, const char *fmtString
, ... );
625 _mesa_vsprintf( char *str
, const char *fmt
, va_list args
);
629 _mesa_warning( __GLcontext
*gc
, const char *fmtString
, ... );
632 _mesa_problem( const __GLcontext
*ctx
, const char *fmtString
, ... );
635 _mesa_error( __GLcontext
*ctx
, GLenum error
, const char *fmtString
, ... );
638 _mesa_debug( const __GLcontext
*ctx
, const char *fmtString
, ... );
645 #endif /* IMPORTS_H */