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 /** Copy \p BYTES bytes from \p SRC into \p DST */
75 #define MEMCPY( DST, SRC, BYTES) _mesa_memcpy(DST, SRC, BYTES)
76 /** Set \p N bytes in \p DST to \p VAL */
77 #define MEMSET( DST, VAL, N ) _mesa_memset(DST, VAL, N)
83 * For GL_ARB_vertex_buffer_object we need to treat vertex array pointers
84 * as offsets into buffer stores. Since the vertex array pointer and
85 * buffer store pointer are both pointers and we need to add them, we use
87 * Both pointers/offsets are expressed in bytes.
89 #define ADD_POINTERS(A, B) ( (GLubyte *) (A) + (uintptr_t) (B) )
93 * Sometimes we treat GLfloats as GLints. On x86 systems, moving a float
94 * as a int (thereby using integer registers instead of FP registers) is
95 * a performance win. Typically, this can be done with ordinary casts.
96 * But with gcc's -fstrict-aliasing flag (which defaults to on in gcc 3.0)
97 * these casts generate warnings.
98 * The following union typedef is used to solve that.
100 typedef union { GLfloat f
; GLint i
; } fi_type
;
104 /**********************************************************************
108 #define MAX_GLUSHORT 0xffff
109 #define MAX_GLUINT 0xffffffff
111 /* Degrees to radians conversion: */
112 #define DEG2RAD (M_PI/180.0)
116 *** SQRTF: single-precision square root
118 #if 0 /* _mesa_sqrtf() not accurate enough - temporarily disabled */
119 # define SQRTF(X) _mesa_sqrtf(X)
121 # define SQRTF(X) (float) sqrt((float) (X))
126 *** INV_SQRTF: single-precision inverse square root
129 #define INV_SQRTF(X) _mesa_inv_sqrt(X)
131 #define INV_SQRTF(X) (1.0F / SQRTF(X)) /* this is faster on a P4 */
136 *** LOG2: Log base 2 of float
140 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
141 * Based on code from http://www.stereopsis.com/log2.html
143 static INLINE GLfloat
LOG2(GLfloat x
)
145 const GLfloat y
= x
* x
* x
* x
;
146 const GLuint ix
= *((GLuint
*) &y
);
147 const GLuint exp
= (ix
>> 23) & 0xFF;
148 const GLint log2
= ((GLint
) exp
) - 127;
149 return (GLfloat
) log2
* (1.0 / 4.0); /* 4, because of x^4 above */
152 /* Pretty fast, and accurate.
153 * Based on code from http://www.flipcode.com/totd/
155 static INLINE GLfloat
LOG2(GLfloat val
)
160 log_2
= ((num
.i
>> 23) & 255) - 128;
161 num
.i
&= ~(255 << 23);
163 num
.f
= ((-1.0f
/3) * num
.f
+ 2) * num
.f
- 2.0f
/3;
164 return num
.f
+ log_2
;
168 * NOTE: log_base_2(x) = log(x) / log(2)
169 * NOTE: 1.442695 = 1/log(2).
171 #define LOG2(x) ((GLfloat) (log(x) * 1.442695F))
176 *** IS_INF_OR_NAN: test if float is infinite or NaN
179 static INLINE
int IS_INF_OR_NAN( float x
)
183 return !(int)((unsigned int)((tmp
.i
& 0x7fffffff)-0x7f800000) >> 31);
185 #elif defined(isfinite)
186 #define IS_INF_OR_NAN(x) (!isfinite(x))
187 #elif defined(finite)
188 #define IS_INF_OR_NAN(x) (!finite(x))
190 #define IS_INF_OR_NAN(x) (!finite(x))
191 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
192 #define IS_INF_OR_NAN(x) (!isfinite(x))
194 #define IS_INF_OR_NAN(x) (!finite(x))
199 *** IS_NEGATIVE: test if float is negative
201 #if defined(USE_IEEE)
202 static INLINE
int GET_FLOAT_BITS( float x
)
208 #define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) < 0)
210 #define IS_NEGATIVE(x) (x < 0.0F)
215 *** DIFFERENT_SIGNS: test if two floats have opposite signs
217 #if defined(USE_IEEE)
218 #define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
220 /* Could just use (x*y<0) except for the flatshading requirements.
221 * Maybe there's a better way?
223 #define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
228 *** CEILF: ceiling of float
229 *** FLOORF: floor of float
230 *** FABSF: absolute value of float
231 *** LOGF: the natural logarithm (base e) of the value
232 *** EXPF: raise e to the value
233 *** LDEXPF: multiply value by an integral power of two
234 *** FREXPF: extract mantissa and exponent from value
236 #if defined(__gnu_linux__)
238 #define CEILF(x) ceilf(x)
239 #define FLOORF(x) floorf(x)
240 #define FABSF(x) fabsf(x)
241 #define LOGF(x) logf(x)
242 #define EXPF(x) expf(x)
243 #define LDEXPF(x,y) ldexpf(x,y)
244 #define FREXPF(x,y) frexpf(x,y)
246 #define CEILF(x) ((GLfloat) ceil(x))
247 #define FLOORF(x) ((GLfloat) floor(x))
248 #define FABSF(x) ((GLfloat) fabs(x))
249 #define LOGF(x) ((GLfloat) log(x))
250 #define EXPF(x) ((GLfloat) exp(x))
251 #define LDEXPF(x,y) ((GLfloat) ldexp(x,y))
252 #define FREXPF(x,y) ((GLfloat) frexp(x,y))
257 *** IROUND: return (as an integer) float rounded to nearest integer
259 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__) && \
260 (!(defined(__BEOS__) || defined(__HAIKU__)) || \
261 (__GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 95)))
262 static INLINE
int iround(float f
)
265 __asm__ ("fistpl %0" : "=m" (r
) : "t" (f
) : "st");
268 #define IROUND(x) iround(x)
269 #elif defined(USE_X86_ASM) && defined(_MSC_VER)
270 static INLINE
int iround(float f
)
279 #define IROUND(x) iround(x)
280 #elif defined(__WATCOMC__) && defined(__386__)
281 long iround(float f
);
282 #pragma aux iround = \
284 "fistp dword ptr [esp]" \
289 #define IROUND(x) iround(x)
291 #define IROUND(f) ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
294 #define IROUND64(f) ((GLint64) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
297 *** IROUND_POS: return (as an integer) positive float rounded to nearest int
300 #define IROUND_POS(f) (assert((f) >= 0.0F), IROUND(f))
302 #define IROUND_POS(f) (IROUND(f))
307 *** IFLOOR: return (as an integer) floor of float
309 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
311 * IEEE floor for computers that round to nearest or even.
312 * 'f' must be between -4194304 and 4194303.
313 * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
314 * but uses some IEEE specific tricks for better speed.
315 * Contributed by Josh Vanderhoof
317 static INLINE
int ifloor(float f
)
321 af
= (3 << 22) + 0.5 + (double)f
;
322 bf
= (3 << 22) + 0.5 - (double)f
;
323 /* GCC generates an extra fstp/fld without this. */
324 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
325 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
326 return (ai
- bi
) >> 1;
328 #define IFLOOR(x) ifloor(x)
329 #elif defined(USE_IEEE)
330 static INLINE
int ifloor(float f
)
336 af
= (3 << 22) + 0.5 + (double)f
;
337 bf
= (3 << 22) + 0.5 - (double)f
;
338 u
.f
= (float) af
; ai
= u
.i
;
339 u
.f
= (float) bf
; bi
= u
.i
;
340 return (ai
- bi
) >> 1;
342 #define IFLOOR(x) ifloor(x)
344 static INLINE
int ifloor(float f
)
347 return (i
> f
) ? i
- 1 : i
;
349 #define IFLOOR(x) ifloor(x)
354 *** ICEIL: return (as an integer) ceiling of float
356 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
358 * IEEE ceil for computers that round to nearest or even.
359 * 'f' must be between -4194304 and 4194303.
360 * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
361 * but uses some IEEE specific tricks for better speed.
362 * Contributed by Josh Vanderhoof
364 static INLINE
int iceil(float f
)
368 af
= (3 << 22) + 0.5 + (double)f
;
369 bf
= (3 << 22) + 0.5 - (double)f
;
370 /* GCC generates an extra fstp/fld without this. */
371 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
372 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
373 return (ai
- bi
+ 1) >> 1;
375 #define ICEIL(x) iceil(x)
376 #elif defined(USE_IEEE)
377 static INLINE
int iceil(float f
)
382 af
= (3 << 22) + 0.5 + (double)f
;
383 bf
= (3 << 22) + 0.5 - (double)f
;
384 u
.f
= (float) af
; ai
= u
.i
;
385 u
.f
= (float) bf
; bi
= u
.i
;
386 return (ai
- bi
+ 1) >> 1;
388 #define ICEIL(x) iceil(x)
390 static INLINE
int iceil(float f
)
393 return (i
< f
) ? i
+ 1 : i
;
395 #define ICEIL(x) iceil(x)
400 * Is x a power of two?
403 _mesa_is_pow_two(int x
)
405 return !(x
& (x
- 1));
409 * Round given integer to next higer power of two
410 * If X is zero result is undefined.
412 * Source for the fallback implementation is
413 * Sean Eron Anderson's webpage "Bit Twiddling Hacks"
414 * http://graphics.stanford.edu/~seander/bithacks.html
416 * When using builtin function have to do some work
417 * for case when passed values 1 to prevent hiting
418 * undefined result from __builtin_clz. Undefined
419 * results would be different depending on optimization
420 * level used for build.
422 static INLINE
int32_t
423 _mesa_next_pow_two_32(uint32_t x
)
426 uint32_t y
= (x
!= 1);
427 return (1 + y
) << ((__builtin_clz(x
- y
) ^ 31) );
440 static INLINE
int64_t
441 _mesa_next_pow_two_64(uint64_t x
)
444 uint64_t y
= (x
!= 1);
445 if (sizeof(x
) == sizeof(long))
446 return (1 + y
) << ((__builtin_clzl(x
- y
) ^ 63));
448 return (1 + y
) << ((__builtin_clzll(x
- y
) ^ 63));
464 *** UNCLAMPED_FLOAT_TO_UBYTE: clamp float to [0,1] and map to ubyte in [0,255]
465 *** CLAMPED_FLOAT_TO_UBYTE: map float known to be in [0,1] to ubyte in [0,255]
467 #if defined(USE_IEEE) && !defined(DEBUG)
468 #define IEEE_0996 0x3f7f0000 /* 0.996 or so */
469 /* This function/macro is sensitive to precision. Test very carefully
472 #define UNCLAMPED_FLOAT_TO_UBYTE(UB, F) \
478 else if (__tmp.i >= IEEE_0996) \
479 UB = (GLubyte) 255; \
481 __tmp.f = __tmp.f * (255.0F/256.0F) + 32768.0F; \
482 UB = (GLubyte) __tmp.i; \
485 #define CLAMPED_FLOAT_TO_UBYTE(UB, F) \
488 __tmp.f = (F) * (255.0F/256.0F) + 32768.0F; \
489 UB = (GLubyte) __tmp.i; \
492 #define UNCLAMPED_FLOAT_TO_UBYTE(ub, f) \
493 ub = ((GLubyte) IROUND(CLAMP((f), 0.0F, 1.0F) * 255.0F))
494 #define CLAMPED_FLOAT_TO_UBYTE(ub, f) \
495 ub = ((GLubyte) IROUND((f) * 255.0F))
500 * Return 1 if this is a little endian machine, 0 if big endian.
502 static INLINE GLboolean
503 _mesa_little_endian(void)
505 const GLuint ui
= 1; /* intentionally not static */
506 return *((const GLubyte
*) &ui
);
511 /**********************************************************************
516 _mesa_malloc( size_t bytes
);
519 _mesa_calloc( size_t bytes
);
522 _mesa_free( void *ptr
);
525 _mesa_align_malloc( size_t bytes
, unsigned long alignment
);
528 _mesa_align_calloc( size_t bytes
, unsigned long alignment
);
531 _mesa_align_free( void *ptr
);
534 _mesa_align_realloc(void *oldBuffer
, size_t oldSize
, size_t newSize
,
535 unsigned long alignment
);
538 _mesa_exec_malloc( GLuint size
);
541 _mesa_exec_free( void *addr
);
544 _mesa_realloc( void *oldBuffer
, size_t oldSize
, size_t newSize
);
547 _mesa_memcpy( void *dest
, const void *src
, size_t n
);
550 _mesa_memset( void *dst
, int val
, size_t n
);
553 _mesa_memset16( unsigned short *dst
, unsigned short val
, size_t n
);
556 _mesa_bzero( void *dst
, size_t n
);
559 _mesa_memcmp( const void *s1
, const void *s2
, size_t n
);
571 _mesa_asinf(float x
);
574 _mesa_atanf(float x
);
577 _mesa_sqrtd(double x
);
580 _mesa_sqrtf(float x
);
583 _mesa_inv_sqrtf(float x
);
586 _mesa_init_sqrt_table(void);
589 _mesa_pow(double x
, double y
);
592 _mesa_ffs(int32_t i
);
595 _mesa_ffsll(int64_t i
);
598 _mesa_bitcount(unsigned int n
);
601 _mesa_float_to_half(float f
);
604 _mesa_half_to_float(GLhalfARB h
);
608 _mesa_bsearch( const void *key
, const void *base
, size_t nmemb
, size_t size
,
609 int (*compar
)(const void *, const void *) );
612 _mesa_getenv( const char *var
);
615 _mesa_strdup( const char *s
);
618 _mesa_strtod( const char *s
, char **end
);
621 _mesa_str_checksum(const char *str
);
624 _mesa_sprintf( char *str
, const char *fmt
, ... );
627 _mesa_snprintf( char *str
, size_t size
, const char *fmt
, ... );
630 _mesa_printf( const char *fmtString
, ... );
633 _mesa_fprintf( FILE *f
, const char *fmtString
, ... );
636 _mesa_vsprintf( char *str
, const char *fmt
, va_list args
);
640 _mesa_warning( __GLcontext
*gc
, const char *fmtString
, ... );
643 _mesa_problem( const __GLcontext
*ctx
, const char *fmtString
, ... );
646 _mesa_error( __GLcontext
*ctx
, GLenum error
, const char *fmtString
, ... );
649 _mesa_debug( const __GLcontext
*ctx
, const char *fmtString
, ... );
656 #endif /* IMPORTS_H */