a9e44a0845ae7cd056be63b8888ab8102f812d6d
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 *** LOG2: Log base 2 of float
124 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
125 * Based on code from http://www.stereopsis.com/log2.html
127 static INLINE GLfloat
LOG2(GLfloat x
)
129 const GLfloat y
= x
* x
* x
* x
;
130 const GLuint ix
= *((GLuint
*) &y
);
131 const GLuint exp
= (ix
>> 23) & 0xFF;
132 const GLint log2
= ((GLint
) exp
) - 127;
133 return (GLfloat
) log2
* (1.0 / 4.0); /* 4, because of x^4 above */
136 /* Pretty fast, and accurate.
137 * Based on code from http://www.flipcode.com/totd/
139 static INLINE GLfloat
LOG2(GLfloat val
)
144 log_2
= ((num
.i
>> 23) & 255) - 128;
145 num
.i
&= ~(255 << 23);
147 num
.f
= ((-1.0f
/3) * num
.f
+ 2) * num
.f
- 2.0f
/3;
148 return num
.f
+ log_2
;
152 * NOTE: log_base_2(x) = log(x) / log(2)
153 * NOTE: 1.442695 = 1/log(2).
155 #define LOG2(x) ((GLfloat) (log(x) * 1.442695F))
160 *** IS_INF_OR_NAN: test if float is infinite or NaN
163 static INLINE
int IS_INF_OR_NAN( float x
)
167 return !(int)((unsigned int)((tmp
.i
& 0x7fffffff)-0x7f800000) >> 31);
169 #elif defined(isfinite)
170 #define IS_INF_OR_NAN(x) (!isfinite(x))
171 #elif defined(finite)
172 #define IS_INF_OR_NAN(x) (!finite(x))
174 #define IS_INF_OR_NAN(x) (!finite(x))
175 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
176 #define IS_INF_OR_NAN(x) (!isfinite(x))
178 #define IS_INF_OR_NAN(x) (!finite(x))
183 *** IS_NEGATIVE: test if float is negative
185 #if defined(USE_IEEE)
186 static INLINE
int GET_FLOAT_BITS( float x
)
192 #define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) < 0)
194 #define IS_NEGATIVE(x) (x < 0.0F)
199 *** DIFFERENT_SIGNS: test if two floats have opposite signs
201 #if defined(USE_IEEE)
202 #define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
204 /* Could just use (x*y<0) except for the flatshading requirements.
205 * Maybe there's a better way?
207 #define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
212 *** CEILF: ceiling of float
213 *** FLOORF: floor of float
214 *** FABSF: absolute value of float
215 *** LOGF: the natural logarithm (base e) of the value
216 *** EXPF: raise e to the value
217 *** LDEXPF: multiply value by an integral power of two
218 *** FREXPF: extract mantissa and exponent from value
220 #if defined(__gnu_linux__)
222 #define CEILF(x) ceilf(x)
223 #define FLOORF(x) floorf(x)
224 #define FABSF(x) fabsf(x)
225 #define LOGF(x) logf(x)
226 #define EXPF(x) expf(x)
227 #define LDEXPF(x,y) ldexpf(x,y)
228 #define FREXPF(x,y) frexpf(x,y)
230 #define CEILF(x) ((GLfloat) ceil(x))
231 #define FLOORF(x) ((GLfloat) floor(x))
232 #define FABSF(x) ((GLfloat) fabs(x))
233 #define LOGF(x) ((GLfloat) log(x))
234 #define EXPF(x) ((GLfloat) exp(x))
235 #define LDEXPF(x,y) ((GLfloat) ldexp(x,y))
236 #define FREXPF(x,y) ((GLfloat) frexp(x,y))
241 *** IROUND: return (as an integer) float rounded to nearest integer
243 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
244 static INLINE
int iround(float f
)
247 __asm__ ("fistpl %0" : "=m" (r
) : "t" (f
) : "st");
250 #define IROUND(x) iround(x)
251 #elif defined(USE_X86_ASM) && defined(_MSC_VER)
252 static INLINE
int iround(float f
)
261 #define IROUND(x) iround(x)
262 #elif defined(__WATCOMC__) && defined(__386__)
263 long iround(float f
);
264 #pragma aux iround = \
266 "fistp dword ptr [esp]" \
271 #define IROUND(x) iround(x)
273 #define IROUND(f) ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
276 #define IROUND64(f) ((GLint64) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
279 *** IROUND_POS: return (as an integer) positive float rounded to nearest int
282 #define IROUND_POS(f) (assert((f) >= 0.0F), IROUND(f))
284 #define IROUND_POS(f) (IROUND(f))
289 *** IFLOOR: return (as an integer) floor of float
291 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
293 * IEEE floor for computers that round to nearest or even.
294 * 'f' must be between -4194304 and 4194303.
295 * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
296 * but uses some IEEE specific tricks for better speed.
297 * Contributed by Josh Vanderhoof
299 static INLINE
int ifloor(float f
)
303 af
= (3 << 22) + 0.5 + (double)f
;
304 bf
= (3 << 22) + 0.5 - (double)f
;
305 /* GCC generates an extra fstp/fld without this. */
306 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
307 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
308 return (ai
- bi
) >> 1;
310 #define IFLOOR(x) ifloor(x)
311 #elif defined(USE_IEEE)
312 static INLINE
int ifloor(float f
)
318 af
= (3 << 22) + 0.5 + (double)f
;
319 bf
= (3 << 22) + 0.5 - (double)f
;
320 u
.f
= (float) af
; ai
= u
.i
;
321 u
.f
= (float) bf
; bi
= u
.i
;
322 return (ai
- bi
) >> 1;
324 #define IFLOOR(x) ifloor(x)
326 static INLINE
int ifloor(float f
)
329 return (i
> f
) ? i
- 1 : i
;
331 #define IFLOOR(x) ifloor(x)
336 *** ICEIL: return (as an integer) ceiling of float
338 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
340 * IEEE ceil for computers that round to nearest or even.
341 * 'f' must be between -4194304 and 4194303.
342 * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
343 * but uses some IEEE specific tricks for better speed.
344 * Contributed by Josh Vanderhoof
346 static INLINE
int iceil(float f
)
350 af
= (3 << 22) + 0.5 + (double)f
;
351 bf
= (3 << 22) + 0.5 - (double)f
;
352 /* GCC generates an extra fstp/fld without this. */
353 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
354 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
355 return (ai
- bi
+ 1) >> 1;
357 #define ICEIL(x) iceil(x)
358 #elif defined(USE_IEEE)
359 static INLINE
int iceil(float f
)
364 af
= (3 << 22) + 0.5 + (double)f
;
365 bf
= (3 << 22) + 0.5 - (double)f
;
366 u
.f
= (float) af
; ai
= u
.i
;
367 u
.f
= (float) bf
; bi
= u
.i
;
368 return (ai
- bi
+ 1) >> 1;
370 #define ICEIL(x) iceil(x)
372 static INLINE
int iceil(float f
)
375 return (i
< f
) ? i
+ 1 : i
;
377 #define ICEIL(x) iceil(x)
382 * Is x a power of two?
385 _mesa_is_pow_two(int x
)
387 return !(x
& (x
- 1));
391 * Round given integer to next higer power of two
392 * If X is zero result is undefined.
394 * Source for the fallback implementation is
395 * Sean Eron Anderson's webpage "Bit Twiddling Hacks"
396 * http://graphics.stanford.edu/~seander/bithacks.html
398 * When using builtin function have to do some work
399 * for case when passed values 1 to prevent hiting
400 * undefined result from __builtin_clz. Undefined
401 * results would be different depending on optimization
402 * level used for build.
404 static INLINE
int32_t
405 _mesa_next_pow_two_32(uint32_t x
)
407 #if defined(__GNUC__) && \
408 ((__GNUC__ == 3 && __GNUC_MINOR__ >= 4) || __GNUC__ >= 4)
409 uint32_t y
= (x
!= 1);
410 return (1 + y
) << ((__builtin_clz(x
- y
) ^ 31) );
423 static INLINE
int64_t
424 _mesa_next_pow_two_64(uint64_t x
)
426 #if defined(__GNUC__) && \
427 ((__GNUC__ == 3 && __GNUC_MINOR__ >= 4) || __GNUC__ >= 4)
428 uint64_t y
= (x
!= 1);
429 if (sizeof(x
) == sizeof(long))
430 return (1 + y
) << ((__builtin_clzl(x
- y
) ^ 63));
432 return (1 + y
) << ((__builtin_clzll(x
- y
) ^ 63));
448 *** UNCLAMPED_FLOAT_TO_UBYTE: clamp float to [0,1] and map to ubyte in [0,255]
449 *** CLAMPED_FLOAT_TO_UBYTE: map float known to be in [0,1] to ubyte in [0,255]
451 #if defined(USE_IEEE) && !defined(DEBUG)
452 #define IEEE_0996 0x3f7f0000 /* 0.996 or so */
453 /* This function/macro is sensitive to precision. Test very carefully
456 #define UNCLAMPED_FLOAT_TO_UBYTE(UB, F) \
462 else if (__tmp.i >= IEEE_0996) \
463 UB = (GLubyte) 255; \
465 __tmp.f = __tmp.f * (255.0F/256.0F) + 32768.0F; \
466 UB = (GLubyte) __tmp.i; \
469 #define CLAMPED_FLOAT_TO_UBYTE(UB, F) \
472 __tmp.f = (F) * (255.0F/256.0F) + 32768.0F; \
473 UB = (GLubyte) __tmp.i; \
476 #define UNCLAMPED_FLOAT_TO_UBYTE(ub, f) \
477 ub = ((GLubyte) IROUND(CLAMP((f), 0.0F, 1.0F) * 255.0F))
478 #define CLAMPED_FLOAT_TO_UBYTE(ub, f) \
479 ub = ((GLubyte) IROUND((f) * 255.0F))
484 * Return 1 if this is a little endian machine, 0 if big endian.
486 static INLINE GLboolean
487 _mesa_little_endian(void)
489 const GLuint ui
= 1; /* intentionally not static */
490 return *((const GLubyte
*) &ui
);
495 /**********************************************************************
500 _mesa_align_malloc( size_t bytes
, unsigned long alignment
);
503 _mesa_align_calloc( size_t bytes
, unsigned long alignment
);
506 _mesa_align_free( void *ptr
);
509 _mesa_align_realloc(void *oldBuffer
, size_t oldSize
, size_t newSize
,
510 unsigned long alignment
);
513 _mesa_exec_malloc( GLuint size
);
516 _mesa_exec_free( void *addr
);
519 _mesa_realloc( void *oldBuffer
, size_t oldSize
, size_t newSize
);
522 _mesa_memset16( unsigned short *dst
, unsigned short val
, size_t n
);
534 _mesa_asinf(float x
);
537 _mesa_atanf(float x
);
540 _mesa_sqrtd(double x
);
543 _mesa_sqrtf(float x
);
546 _mesa_inv_sqrtf(float x
);
549 _mesa_init_sqrt_table(void);
552 _mesa_ffs(int32_t i
);
555 _mesa_ffsll(int64_t i
);
558 _mesa_bitcount(unsigned int n
);
561 _mesa_float_to_half(float f
);
564 _mesa_half_to_float(GLhalfARB h
);
568 _mesa_bsearch( const void *key
, const void *base
, size_t nmemb
, size_t size
,
569 int (*compar
)(const void *, const void *) );
572 _mesa_getenv( const char *var
);
575 _mesa_strdup( const char *s
);
578 _mesa_strtof( const char *s
, char **end
);
581 _mesa_str_checksum(const char *str
);
584 _mesa_snprintf( char *str
, size_t size
, const char *fmt
, ... );
587 _mesa_warning( __GLcontext
*gc
, const char *fmtString
, ... );
590 _mesa_problem( const __GLcontext
*ctx
, const char *fmtString
, ... );
593 _mesa_error( __GLcontext
*ctx
, GLenum error
, const char *fmtString
, ... );
596 _mesa_debug( const __GLcontext
*ctx
, const char *fmtString
, ... );
603 #endif /* IMPORTS_H */