4eabdfdb0d249173d4f35daaa4318da7fc50a4d2
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)
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)
78 * For GL_ARB_vertex_buffer_object we need to treat vertex array pointers
79 * as offsets into buffer stores. Since the vertex array pointer and
80 * buffer store pointer are both pointers and we need to add them, we use
82 * Both pointers/offsets are expressed in bytes.
84 #define ADD_POINTERS(A, B) ( (GLubyte *) (A) + (uintptr_t) (B) )
88 * Sometimes we treat GLfloats as GLints. On x86 systems, moving a float
89 * as a int (thereby using integer registers instead of FP registers) is
90 * a performance win. Typically, this can be done with ordinary casts.
91 * But with gcc's -fstrict-aliasing flag (which defaults to on in gcc 3.0)
92 * these casts generate warnings.
93 * The following union typedef is used to solve that.
95 typedef union { GLfloat f
; GLint i
; } fi_type
;
99 /**********************************************************************
103 #define MAX_GLUSHORT 0xffff
104 #define MAX_GLUINT 0xffffffff
106 /* Degrees to radians conversion: */
107 #define DEG2RAD (M_PI/180.0)
111 *** SQRTF: single-precision square root
113 #if 0 /* _mesa_sqrtf() not accurate enough - temporarily disabled */
114 # define SQRTF(X) _mesa_sqrtf(X)
116 # define SQRTF(X) (float) sqrt((float) (X))
121 *** INV_SQRTF: single-precision inverse square root
124 #define INV_SQRTF(X) _mesa_inv_sqrt(X)
126 #define INV_SQRTF(X) (1.0F / SQRTF(X)) /* this is faster on a P4 */
131 *** LOG2: Log base 2 of float
135 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
136 * Based on code from http://www.stereopsis.com/log2.html
138 static INLINE GLfloat
LOG2(GLfloat x
)
140 const GLfloat y
= x
* x
* x
* x
;
141 const GLuint ix
= *((GLuint
*) &y
);
142 const GLuint exp
= (ix
>> 23) & 0xFF;
143 const GLint log2
= ((GLint
) exp
) - 127;
144 return (GLfloat
) log2
* (1.0 / 4.0); /* 4, because of x^4 above */
147 /* Pretty fast, and accurate.
148 * Based on code from http://www.flipcode.com/totd/
150 static INLINE GLfloat
LOG2(GLfloat val
)
155 log_2
= ((num
.i
>> 23) & 255) - 128;
156 num
.i
&= ~(255 << 23);
158 num
.f
= ((-1.0f
/3) * num
.f
+ 2) * num
.f
- 2.0f
/3;
159 return num
.f
+ log_2
;
163 * NOTE: log_base_2(x) = log(x) / log(2)
164 * NOTE: 1.442695 = 1/log(2).
166 #define LOG2(x) ((GLfloat) (log(x) * 1.442695F))
171 *** IS_INF_OR_NAN: test if float is infinite or NaN
174 static INLINE
int IS_INF_OR_NAN( float x
)
178 return !(int)((unsigned int)((tmp
.i
& 0x7fffffff)-0x7f800000) >> 31);
180 #elif defined(isfinite)
181 #define IS_INF_OR_NAN(x) (!isfinite(x))
182 #elif defined(finite)
183 #define IS_INF_OR_NAN(x) (!finite(x))
185 #define IS_INF_OR_NAN(x) (!finite(x))
186 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
187 #define IS_INF_OR_NAN(x) (!isfinite(x))
189 #define IS_INF_OR_NAN(x) (!finite(x))
194 *** IS_NEGATIVE: test if float is negative
196 #if defined(USE_IEEE)
197 static INLINE
int GET_FLOAT_BITS( float x
)
203 #define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) < 0)
205 #define IS_NEGATIVE(x) (x < 0.0F)
210 *** DIFFERENT_SIGNS: test if two floats have opposite signs
212 #if defined(USE_IEEE)
213 #define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
215 /* Could just use (x*y<0) except for the flatshading requirements.
216 * Maybe there's a better way?
218 #define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
223 *** CEILF: ceiling of float
224 *** FLOORF: floor of float
225 *** FABSF: absolute value of float
226 *** LOGF: the natural logarithm (base e) of the value
227 *** EXPF: raise e to the value
228 *** LDEXPF: multiply value by an integral power of two
229 *** FREXPF: extract mantissa and exponent from value
231 #if defined(__gnu_linux__)
233 #define CEILF(x) ceilf(x)
234 #define FLOORF(x) floorf(x)
235 #define FABSF(x) fabsf(x)
236 #define LOGF(x) logf(x)
237 #define EXPF(x) expf(x)
238 #define LDEXPF(x,y) ldexpf(x,y)
239 #define FREXPF(x,y) frexpf(x,y)
241 #define CEILF(x) ((GLfloat) ceil(x))
242 #define FLOORF(x) ((GLfloat) floor(x))
243 #define FABSF(x) ((GLfloat) fabs(x))
244 #define LOGF(x) ((GLfloat) log(x))
245 #define EXPF(x) ((GLfloat) exp(x))
246 #define LDEXPF(x,y) ((GLfloat) ldexp(x,y))
247 #define FREXPF(x,y) ((GLfloat) frexp(x,y))
252 *** IROUND: return (as an integer) float rounded to nearest integer
254 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__) && \
255 (!(defined(__BEOS__) || defined(__HAIKU__)) || \
256 (__GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 95)))
257 static INLINE
int iround(float f
)
260 __asm__ ("fistpl %0" : "=m" (r
) : "t" (f
) : "st");
263 #define IROUND(x) iround(x)
264 #elif defined(USE_X86_ASM) && defined(_MSC_VER)
265 static INLINE
int iround(float f
)
274 #define IROUND(x) iround(x)
275 #elif defined(__WATCOMC__) && defined(__386__)
276 long iround(float f
);
277 #pragma aux iround = \
279 "fistp dword ptr [esp]" \
284 #define IROUND(x) iround(x)
286 #define IROUND(f) ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
289 #define IROUND64(f) ((GLint64) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
292 *** IROUND_POS: return (as an integer) positive float rounded to nearest int
295 #define IROUND_POS(f) (assert((f) >= 0.0F), IROUND(f))
297 #define IROUND_POS(f) (IROUND(f))
302 *** IFLOOR: return (as an integer) floor of float
304 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
306 * IEEE floor for computers that round to nearest or even.
307 * 'f' must be between -4194304 and 4194303.
308 * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
309 * but uses some IEEE specific tricks for better speed.
310 * Contributed by Josh Vanderhoof
312 static INLINE
int ifloor(float f
)
316 af
= (3 << 22) + 0.5 + (double)f
;
317 bf
= (3 << 22) + 0.5 - (double)f
;
318 /* GCC generates an extra fstp/fld without this. */
319 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
320 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
321 return (ai
- bi
) >> 1;
323 #define IFLOOR(x) ifloor(x)
324 #elif defined(USE_IEEE)
325 static INLINE
int ifloor(float f
)
331 af
= (3 << 22) + 0.5 + (double)f
;
332 bf
= (3 << 22) + 0.5 - (double)f
;
333 u
.f
= (float) af
; ai
= u
.i
;
334 u
.f
= (float) bf
; bi
= u
.i
;
335 return (ai
- bi
) >> 1;
337 #define IFLOOR(x) ifloor(x)
339 static INLINE
int ifloor(float f
)
342 return (i
> f
) ? i
- 1 : i
;
344 #define IFLOOR(x) ifloor(x)
349 *** ICEIL: return (as an integer) ceiling of float
351 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
353 * IEEE ceil for computers that round to nearest or even.
354 * 'f' must be between -4194304 and 4194303.
355 * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
356 * but uses some IEEE specific tricks for better speed.
357 * Contributed by Josh Vanderhoof
359 static INLINE
int iceil(float f
)
363 af
= (3 << 22) + 0.5 + (double)f
;
364 bf
= (3 << 22) + 0.5 - (double)f
;
365 /* GCC generates an extra fstp/fld without this. */
366 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
367 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
368 return (ai
- bi
+ 1) >> 1;
370 #define ICEIL(x) iceil(x)
371 #elif defined(USE_IEEE)
372 static INLINE
int iceil(float f
)
377 af
= (3 << 22) + 0.5 + (double)f
;
378 bf
= (3 << 22) + 0.5 - (double)f
;
379 u
.f
= (float) af
; ai
= u
.i
;
380 u
.f
= (float) bf
; bi
= u
.i
;
381 return (ai
- bi
+ 1) >> 1;
383 #define ICEIL(x) iceil(x)
385 static INLINE
int iceil(float f
)
388 return (i
< f
) ? i
+ 1 : i
;
390 #define ICEIL(x) iceil(x)
395 * Is x a power of two?
398 _mesa_is_pow_two(int x
)
400 return !(x
& (x
- 1));
404 * Round given integer to next higer power of two
405 * If X is zero result is undefined.
407 * Source for the fallback implementation is
408 * Sean Eron Anderson's webpage "Bit Twiddling Hacks"
409 * http://graphics.stanford.edu/~seander/bithacks.html
411 * When using builtin function have to do some work
412 * for case when passed values 1 to prevent hiting
413 * undefined result from __builtin_clz. Undefined
414 * results would be different depending on optimization
415 * level used for build.
417 static INLINE
int32_t
418 _mesa_next_pow_two_32(uint32_t x
)
421 uint32_t y
= (x
!= 1);
422 return (1 + y
) << ((__builtin_clz(x
- y
) ^ 31) );
435 static INLINE
int64_t
436 _mesa_next_pow_two_64(uint64_t x
)
439 uint64_t y
= (x
!= 1);
440 if (sizeof(x
) == sizeof(long))
441 return (1 + y
) << ((__builtin_clzl(x
- y
) ^ 63));
443 return (1 + y
) << ((__builtin_clzll(x
- y
) ^ 63));
459 *** UNCLAMPED_FLOAT_TO_UBYTE: clamp float to [0,1] and map to ubyte in [0,255]
460 *** CLAMPED_FLOAT_TO_UBYTE: map float known to be in [0,1] to ubyte in [0,255]
462 #if defined(USE_IEEE) && !defined(DEBUG)
463 #define IEEE_0996 0x3f7f0000 /* 0.996 or so */
464 /* This function/macro is sensitive to precision. Test very carefully
467 #define UNCLAMPED_FLOAT_TO_UBYTE(UB, F) \
473 else if (__tmp.i >= IEEE_0996) \
474 UB = (GLubyte) 255; \
476 __tmp.f = __tmp.f * (255.0F/256.0F) + 32768.0F; \
477 UB = (GLubyte) __tmp.i; \
480 #define CLAMPED_FLOAT_TO_UBYTE(UB, F) \
483 __tmp.f = (F) * (255.0F/256.0F) + 32768.0F; \
484 UB = (GLubyte) __tmp.i; \
487 #define UNCLAMPED_FLOAT_TO_UBYTE(ub, f) \
488 ub = ((GLubyte) IROUND(CLAMP((f), 0.0F, 1.0F) * 255.0F))
489 #define CLAMPED_FLOAT_TO_UBYTE(ub, f) \
490 ub = ((GLubyte) IROUND((f) * 255.0F))
495 * Return 1 if this is a little endian machine, 0 if big endian.
497 static INLINE GLboolean
498 _mesa_little_endian(void)
500 const GLuint ui
= 1; /* intentionally not static */
501 return *((const GLubyte
*) &ui
);
506 /**********************************************************************
511 _mesa_align_malloc( size_t bytes
, unsigned long alignment
);
514 _mesa_align_calloc( size_t bytes
, unsigned long alignment
);
517 _mesa_align_free( void *ptr
);
520 _mesa_align_realloc(void *oldBuffer
, size_t oldSize
, size_t newSize
,
521 unsigned long alignment
);
524 _mesa_exec_malloc( GLuint size
);
527 _mesa_exec_free( void *addr
);
530 _mesa_realloc( void *oldBuffer
, size_t oldSize
, size_t newSize
);
533 _mesa_memset16( unsigned short *dst
, unsigned short val
, size_t n
);
545 _mesa_asinf(float x
);
548 _mesa_atanf(float x
);
551 _mesa_sqrtd(double x
);
554 _mesa_sqrtf(float x
);
557 _mesa_inv_sqrtf(float x
);
560 _mesa_init_sqrt_table(void);
563 _mesa_pow(double x
, double y
);
566 _mesa_ffs(int32_t i
);
569 _mesa_ffsll(int64_t i
);
572 _mesa_bitcount(unsigned int n
);
575 _mesa_float_to_half(float f
);
578 _mesa_half_to_float(GLhalfARB h
);
582 _mesa_bsearch( const void *key
, const void *base
, size_t nmemb
, size_t size
,
583 int (*compar
)(const void *, const void *) );
586 _mesa_getenv( const char *var
);
589 _mesa_strdup( const char *s
);
592 _mesa_strtod( const char *s
, char **end
);
595 _mesa_str_checksum(const char *str
);
598 _mesa_sprintf( char *str
, const char *fmt
, ... );
601 _mesa_snprintf( char *str
, size_t size
, const char *fmt
, ... );
604 _mesa_printf( const char *fmtString
, ... );
607 _mesa_fprintf( FILE *f
, const char *fmtString
, ... );
610 _mesa_vsprintf( char *str
, const char *fmt
, va_list args
);
614 _mesa_warning( __GLcontext
*gc
, const char *fmtString
, ... );
617 _mesa_problem( const __GLcontext
*ctx
, const char *fmtString
, ... );
620 _mesa_error( __GLcontext
*ctx
, GLenum error
, const char *fmtString
, ... );
623 _mesa_debug( const __GLcontext
*ctx
, const char *fmtString
, ... );
630 #endif /* IMPORTS_H */