bcbcd0f3e27bf2aaf015c6295d71edc427bd1f4b
2 * Mesa 3-D graphics library
4 * Copyright (C) 1999-2008 Brian Paul All Rights Reserved.
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the "Software"),
8 * to deal in the Software without restriction, including without limitation
9 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
10 * and/or sell copies of the Software, and to permit persons to whom the
11 * Software is furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included
14 * in all copies or substantial portions of the Software.
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
17 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
20 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
21 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
22 * 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 a structure of type \p T */
53 #define MALLOC_STRUCT(T) (struct T *) malloc(sizeof(struct T))
54 /** Allocate and zero a structure of type \p T */
55 #define CALLOC_STRUCT(T) (struct T *) calloc(1, sizeof(struct T))
61 * For GL_ARB_vertex_buffer_object we need to treat vertex array pointers
62 * as offsets into buffer stores. Since the vertex array pointer and
63 * buffer store pointer are both pointers and we need to add them, we use
65 * Both pointers/offsets are expressed in bytes.
67 #define ADD_POINTERS(A, B) ( (GLubyte *) (A) + (uintptr_t) (B) )
71 * Sometimes we treat GLfloats as GLints. On x86 systems, moving a float
72 * as a int (thereby using integer registers instead of FP registers) is
73 * a performance win. Typically, this can be done with ordinary casts.
74 * But with gcc's -fstrict-aliasing flag (which defaults to on in gcc 3.0)
75 * these casts generate warnings.
76 * The following union typedef is used to solve that.
78 typedef union { GLfloat f
; GLint i
; GLuint u
; } fi_type
;
82 /**********************************************************************
86 #define MAX_GLUSHORT 0xffff
87 #define MAX_GLUINT 0xffffffff
89 /* Degrees to radians conversion: */
90 #define DEG2RAD (M_PI/180.0)
94 * \name Work-arounds for platforms that lack C99 math functions
97 #if (!defined(_XOPEN_SOURCE) || (_XOPEN_SOURCE < 600)) && !defined(_ISOC99_SOURCE) \
98 && (!defined(__STDC_VERSION__) || (__STDC_VERSION__ < 199901L)) \
99 && (!defined(_MSC_VER) || (_MSC_VER < 1400))
100 #define acosf(f) ((float) acos(f))
101 #define asinf(f) ((float) asin(f))
102 #define atan2f(x,y) ((float) atan2(x,y))
103 #define atanf(f) ((float) atan(f))
104 #define ceilf(f) ((float) ceil(f))
105 #define cosf(f) ((float) cos(f))
106 #define coshf(f) ((float) cosh(f))
107 #define expf(f) ((float) exp(f))
108 #define exp2f(f) ((float) exp2(f))
109 #define floorf(f) ((float) floor(f))
110 #define logf(f) ((float) log(f))
113 #define log2f(f) (logf(f) * (float) (1.0 / M_LN2))
115 #define log2f(f) ((float) log2(f))
118 #define powf(x,y) ((float) pow(x,y))
119 #define sinf(f) ((float) sin(f))
120 #define sinhf(f) ((float) sinh(f))
121 #define sqrtf(f) ((float) sqrt(f))
122 #define tanf(f) ((float) tan(f))
123 #define tanhf(f) ((float) tanh(f))
124 #define acoshf(f) ((float) acosh(f))
125 #define asinhf(f) ((float) asinh(f))
126 #define atanhf(f) ((float) atanh(f))
129 #if defined(_MSC_VER)
130 #if _MSC_VER < 1800 /* Not req'd on VS2013 and above */
131 static inline float truncf(float x
) { return x
< 0.0f
? ceilf(x
) : floorf(x
); }
132 static inline float exp2f(float x
) { return powf(2.0f
, x
); }
133 static inline float log2f(float x
) { return logf(x
) * 1.442695041f
; }
134 static inline float asinhf(float x
) { return logf(x
+ sqrtf(x
* x
+ 1.0f
)); }
135 static inline float acoshf(float x
) { return logf(x
+ sqrtf(x
* x
- 1.0f
)); }
136 static inline float atanhf(float x
) { return (logf(1.0f
+ x
) - logf(1.0f
- x
)) / 2.0f
; }
137 static inline int isblank(int ch
) { return ch
== ' ' || ch
== '\t'; }
138 #define strtoll(p, e, b) _strtoi64(p, e, b)
139 #endif /* _MSC_VER < 1800 */
140 #define strcasecmp(s1, s2) _stricmp(s1, s2)
146 * signbit() is a macro on Linux. Not available on Windows.
149 #define signbit(x) ((x) < 0.0f)
153 /** single-precision inverse square root */
157 /* XXX we could try Quake's fast inverse square root function here */
158 return 1.0F
/ sqrtf(x
);
163 *** LOG2: Log base 2 of float
165 static inline GLfloat
LOG2(GLfloat x
)
168 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
169 * Based on code from http://www.stereopsis.com/log2.html
171 const GLfloat y
= x
* x
* x
* x
;
172 const GLuint ix
= *((GLuint
*) &y
);
173 const GLuint exp
= (ix
>> 23) & 0xFF;
174 const GLint log2
= ((GLint
) exp
) - 127;
175 return (GLfloat
) log2
* (1.0 / 4.0); /* 4, because of x^4 above */
177 /* Pretty fast, and accurate.
178 * Based on code from http://www.flipcode.com/totd/
183 log_2
= ((num
.i
>> 23) & 255) - 128;
184 num
.i
&= ~(255 << 23);
186 num
.f
= ((-1.0f
/3) * num
.f
+ 2) * num
.f
- 2.0f
/3;
187 return num
.f
+ log_2
;
193 *** IS_INF_OR_NAN: test if float is infinite or NaN
195 #if defined(isfinite)
196 #define IS_INF_OR_NAN(x) (!isfinite(x))
197 #elif defined(finite)
198 #define IS_INF_OR_NAN(x) (!finite(x))
199 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
200 #define IS_INF_OR_NAN(x) (!isfinite(x))
202 #define IS_INF_OR_NAN(x) (!finite(x))
207 *** FLOORF: floor of float
208 *** FABSF: absolute value of float
210 #if defined(__gnu_linux__)
212 #define FLOORF(x) floorf(x)
213 #define FABSF(x) fabsf(x)
215 #define FLOORF(x) ((GLfloat) floor(x))
216 #define FABSF(x) ((GLfloat) fabs(x))
221 * Convert float to int by rounding to nearest integer, away from zero.
223 static inline int IROUND(float f
)
225 return (int) ((f
>= 0.0F
) ? (f
+ 0.5F
) : (f
- 0.5F
));
230 * Convert float to int64 by rounding to nearest integer.
232 static inline GLint64
IROUND64(float f
)
234 return (GLint64
) ((f
>= 0.0F
) ? (f
+ 0.5F
) : (f
- 0.5F
));
239 * Convert positive float to int by rounding to nearest integer.
241 static inline int IROUND_POS(float f
)
244 return (int) (f
+ 0.5F
);
248 # include <xmmintrin.h>
252 * Convert float to int using a fast method. The rounding mode may vary.
254 static inline int F_TO_I(float f
)
256 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
258 __asm__ ("fistpl %0" : "=m" (r
) : "t" (f
) : "st");
260 #elif defined(USE_X86_ASM) && defined(_MSC_VER)
267 #elif defined(__x86_64__)
268 return _mm_cvt_ss2si(_mm_load_ss(&f
));
275 /** Return (as an integer) floor of float */
276 static inline int IFLOOR(float f
)
278 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
280 * IEEE floor for computers that round to nearest or even.
281 * 'f' must be between -4194304 and 4194303.
282 * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
283 * but uses some IEEE specific tricks for better speed.
284 * Contributed by Josh Vanderhoof
288 af
= (3 << 22) + 0.5 + (double)f
;
289 bf
= (3 << 22) + 0.5 - (double)f
;
290 /* GCC generates an extra fstp/fld without this. */
291 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
292 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
293 return (ai
- bi
) >> 1;
298 af
= (3 << 22) + 0.5 + (double)f
;
299 bf
= (3 << 22) + 0.5 - (double)f
;
300 u
.f
= (float) af
; ai
= u
.i
;
301 u
.f
= (float) bf
; bi
= u
.i
;
302 return (ai
- bi
) >> 1;
307 /** Return (as an integer) ceiling of float */
308 static inline int ICEIL(float f
)
310 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
312 * IEEE ceil for computers that round to nearest or even.
313 * 'f' must be between -4194304 and 4194303.
314 * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
315 * but uses some IEEE specific tricks for better speed.
316 * Contributed by Josh Vanderhoof
320 af
= (3 << 22) + 0.5 + (double)f
;
321 bf
= (3 << 22) + 0.5 - (double)f
;
322 /* GCC generates an extra fstp/fld without this. */
323 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
324 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
325 return (ai
- bi
+ 1) >> 1;
330 af
= (3 << 22) + 0.5 + (double)f
;
331 bf
= (3 << 22) + 0.5 - (double)f
;
332 u
.f
= (float) af
; ai
= u
.i
;
333 u
.f
= (float) bf
; bi
= u
.i
;
334 return (ai
- bi
+ 1) >> 1;
340 * Is x a power of two?
343 _mesa_is_pow_two(int x
)
345 return !(x
& (x
- 1));
349 * Round given integer to next higer power of two
350 * If X is zero result is undefined.
352 * Source for the fallback implementation is
353 * Sean Eron Anderson's webpage "Bit Twiddling Hacks"
354 * http://graphics.stanford.edu/~seander/bithacks.html
356 * When using builtin function have to do some work
357 * for case when passed values 1 to prevent hiting
358 * undefined result from __builtin_clz. Undefined
359 * results would be different depending on optimization
360 * level used for build.
362 static inline int32_t
363 _mesa_next_pow_two_32(uint32_t x
)
365 #ifdef HAVE___BUILTIN_CLZ
366 uint32_t y
= (x
!= 1);
367 return (1 + y
) << ((__builtin_clz(x
- y
) ^ 31) );
380 static inline int64_t
381 _mesa_next_pow_two_64(uint64_t x
)
383 #ifdef HAVE___BUILTIN_CLZLL
384 uint64_t y
= (x
!= 1);
385 STATIC_ASSERT(sizeof(x
) == sizeof(long long));
386 return (1 + y
) << ((__builtin_clzll(x
- y
) ^ 63));
402 * Returns the floor form of binary logarithm for a 32-bit integer.
405 _mesa_logbase2(GLuint n
)
407 #ifdef HAVE___BUILTIN_CLZ
408 return (31 - __builtin_clz(n
| 1));
411 if (n
>= 1<<16) { n
>>= 16; pos
+= 16; }
412 if (n
>= 1<< 8) { n
>>= 8; pos
+= 8; }
413 if (n
>= 1<< 4) { n
>>= 4; pos
+= 4; }
414 if (n
>= 1<< 2) { n
>>= 2; pos
+= 2; }
415 if (n
>= 1<< 1) { pos
+= 1; }
422 * Return 1 if this is a little endian machine, 0 if big endian.
424 static inline GLboolean
425 _mesa_little_endian(void)
427 const GLuint ui
= 1; /* intentionally not static */
428 return *((const GLubyte
*) &ui
);
433 /**********************************************************************
438 _mesa_align_malloc( size_t bytes
, unsigned long alignment
);
441 _mesa_align_calloc( size_t bytes
, unsigned long alignment
);
444 _mesa_align_free( void *ptr
);
447 _mesa_align_realloc(void *oldBuffer
, size_t oldSize
, size_t newSize
,
448 unsigned long alignment
);
451 _mesa_exec_malloc( GLuint size
);
454 _mesa_exec_free( void *addr
);
458 #define FFS_DEFINED 1
459 #ifdef HAVE___BUILTIN_FFS
460 #define ffs __builtin_ffs
462 extern int ffs(int i
);
465 #ifdef HAVE___BUILTIN_FFSLL
466 #define ffsll __builtin_ffsll
468 extern int ffsll(long long int i
);
470 #endif /* FFS_DEFINED */
473 #ifdef HAVE___BUILTIN_POPCOUNT
474 #define _mesa_bitcount(i) __builtin_popcount(i)
477 _mesa_bitcount(unsigned int n
);
480 #ifdef HAVE___BUILTIN_POPCOUNTLL
481 #define _mesa_bitcount_64(i) __builtin_popcountll(i)
484 _mesa_bitcount_64(uint64_t n
);
488 * Find the last (most significant) bit set in a word.
490 * Essentially ffs() in the reverse direction.
492 static inline unsigned int
493 _mesa_fls(unsigned int n
)
495 #ifdef HAVE___BUILTIN_CLZ
496 return n
== 0 ? 0 : 32 - __builtin_clz(n
);
511 _mesa_round_to_even(float val
);
514 _mesa_float_to_half(float f
);
517 _mesa_half_to_float(GLhalfARB h
);
520 _mesa_half_is_negative(GLhalfARB h
)
526 _mesa_strdup( const char *s
);
529 _mesa_str_checksum(const char *str
);
532 _mesa_snprintf( char *str
, size_t size
, const char *fmt
, ... ) PRINTFLIKE(3, 4);
535 _mesa_vsnprintf(char *str
, size_t size
, const char *fmt
, va_list arg
);
538 #if defined(_MSC_VER) && !defined(snprintf)
539 #define snprintf _snprintf
548 #endif /* IMPORTS_H */