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 #if _MSC_VER < 1800 /* Not req'd on VS2013 and above */
95 static inline int isblank(int ch
) { return ch
== ' ' || ch
== '\t'; }
96 #define strtoll(p, e, b) _strtoi64(p, e, b)
97 #endif /* _MSC_VER < 1800 */
98 #define strcasecmp(s1, s2) _stricmp(s1, s2)
104 * signbit() is a macro on Linux. Not available on Windows.
107 #define signbit(x) ((x) < 0.0f)
112 *** LOG2: Log base 2 of float
114 static inline GLfloat
LOG2(GLfloat x
)
117 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
118 * Based on code from http://www.stereopsis.com/log2.html
120 const GLfloat y
= x
* x
* x
* x
;
121 const GLuint ix
= *((GLuint
*) &y
);
122 const GLuint exp
= (ix
>> 23) & 0xFF;
123 const GLint log2
= ((GLint
) exp
) - 127;
124 return (GLfloat
) log2
* (1.0 / 4.0); /* 4, because of x^4 above */
126 /* Pretty fast, and accurate.
127 * Based on code from http://www.flipcode.com/totd/
132 log_2
= ((num
.i
>> 23) & 255) - 128;
133 num
.i
&= ~(255 << 23);
135 num
.f
= ((-1.0f
/3) * num
.f
+ 2) * num
.f
- 2.0f
/3;
136 return num
.f
+ log_2
;
142 *** IS_INF_OR_NAN: test if float is infinite or NaN
144 #if defined(isfinite)
145 #define IS_INF_OR_NAN(x) (!isfinite(x))
146 #elif defined(finite)
147 #define IS_INF_OR_NAN(x) (!finite(x))
148 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
149 #define IS_INF_OR_NAN(x) (!isfinite(x))
151 #define IS_INF_OR_NAN(x) (!finite(x))
156 * Convert float to int by rounding to nearest integer, away from zero.
158 static inline int IROUND(float f
)
160 return (int) ((f
>= 0.0F
) ? (f
+ 0.5F
) : (f
- 0.5F
));
165 * Convert float to int64 by rounding to nearest integer.
167 static inline GLint64
IROUND64(float f
)
169 return (GLint64
) ((f
>= 0.0F
) ? (f
+ 0.5F
) : (f
- 0.5F
));
174 * Convert positive float to int by rounding to nearest integer.
176 static inline int IROUND_POS(float f
)
179 return (int) (f
+ 0.5F
);
183 # include <xmmintrin.h>
187 * Convert float to int using a fast method. The rounding mode may vary.
189 static inline int F_TO_I(float f
)
191 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
193 __asm__ ("fistpl %0" : "=m" (r
) : "t" (f
) : "st");
195 #elif defined(USE_X86_ASM) && defined(_MSC_VER)
202 #elif defined(__x86_64__)
203 return _mm_cvt_ss2si(_mm_load_ss(&f
));
210 /** Return (as an integer) floor of float */
211 static inline int IFLOOR(float f
)
213 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
215 * IEEE floor for computers that round to nearest or even.
216 * 'f' must be between -4194304 and 4194303.
217 * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
218 * but uses some IEEE specific tricks for better speed.
219 * Contributed by Josh Vanderhoof
223 af
= (3 << 22) + 0.5 + (double)f
;
224 bf
= (3 << 22) + 0.5 - (double)f
;
225 /* GCC generates an extra fstp/fld without this. */
226 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
227 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
228 return (ai
- bi
) >> 1;
233 af
= (3 << 22) + 0.5 + (double)f
;
234 bf
= (3 << 22) + 0.5 - (double)f
;
235 u
.f
= (float) af
; ai
= u
.i
;
236 u
.f
= (float) bf
; bi
= u
.i
;
237 return (ai
- bi
) >> 1;
242 /** Return (as an integer) ceiling of float */
243 static inline int ICEIL(float f
)
245 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
247 * IEEE ceil for computers that round to nearest or even.
248 * 'f' must be between -4194304 and 4194303.
249 * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
250 * but uses some IEEE specific tricks for better speed.
251 * Contributed by Josh Vanderhoof
255 af
= (3 << 22) + 0.5 + (double)f
;
256 bf
= (3 << 22) + 0.5 - (double)f
;
257 /* GCC generates an extra fstp/fld without this. */
258 __asm__ ("fstps %0" : "=m" (ai
) : "t" (af
) : "st");
259 __asm__ ("fstps %0" : "=m" (bi
) : "t" (bf
) : "st");
260 return (ai
- bi
+ 1) >> 1;
265 af
= (3 << 22) + 0.5 + (double)f
;
266 bf
= (3 << 22) + 0.5 - (double)f
;
267 u
.f
= (float) af
; ai
= u
.i
;
268 u
.f
= (float) bf
; bi
= u
.i
;
269 return (ai
- bi
+ 1) >> 1;
275 * Is x a power of two?
278 _mesa_is_pow_two(int x
)
280 return !(x
& (x
- 1));
284 * Round given integer to next higer power of two
285 * If X is zero result is undefined.
287 * Source for the fallback implementation is
288 * Sean Eron Anderson's webpage "Bit Twiddling Hacks"
289 * http://graphics.stanford.edu/~seander/bithacks.html
291 * When using builtin function have to do some work
292 * for case when passed values 1 to prevent hiting
293 * undefined result from __builtin_clz. Undefined
294 * results would be different depending on optimization
295 * level used for build.
297 static inline int32_t
298 _mesa_next_pow_two_32(uint32_t x
)
300 #ifdef HAVE___BUILTIN_CLZ
301 uint32_t y
= (x
!= 1);
302 return (1 + y
) << ((__builtin_clz(x
- y
) ^ 31) );
315 static inline int64_t
316 _mesa_next_pow_two_64(uint64_t x
)
318 #ifdef HAVE___BUILTIN_CLZLL
319 uint64_t y
= (x
!= 1);
320 STATIC_ASSERT(sizeof(x
) == sizeof(long long));
321 return (1 + y
) << ((__builtin_clzll(x
- y
) ^ 63));
337 * Returns the floor form of binary logarithm for a 32-bit integer.
340 _mesa_logbase2(GLuint n
)
342 #ifdef HAVE___BUILTIN_CLZ
343 return (31 - __builtin_clz(n
| 1));
346 if (n
>= 1<<16) { n
>>= 16; pos
+= 16; }
347 if (n
>= 1<< 8) { n
>>= 8; pos
+= 8; }
348 if (n
>= 1<< 4) { n
>>= 4; pos
+= 4; }
349 if (n
>= 1<< 2) { n
>>= 2; pos
+= 2; }
350 if (n
>= 1<< 1) { pos
+= 1; }
357 * Return 1 if this is a little endian machine, 0 if big endian.
359 static inline GLboolean
360 _mesa_little_endian(void)
362 const GLuint ui
= 1; /* intentionally not static */
363 return *((const GLubyte
*) &ui
);
368 /**********************************************************************
373 _mesa_align_malloc( size_t bytes
, unsigned long alignment
);
376 _mesa_align_calloc( size_t bytes
, unsigned long alignment
);
379 _mesa_align_free( void *ptr
);
382 _mesa_align_realloc(void *oldBuffer
, size_t oldSize
, size_t newSize
,
383 unsigned long alignment
);
386 _mesa_exec_malloc( GLuint size
);
389 _mesa_exec_free( void *addr
);
393 #define FFS_DEFINED 1
394 #ifdef HAVE___BUILTIN_FFS
395 #define ffs __builtin_ffs
397 extern int ffs(int i
);
400 #ifdef HAVE___BUILTIN_FFSLL
401 #define ffsll __builtin_ffsll
403 extern int ffsll(long long int i
);
405 #endif /* FFS_DEFINED */
408 #ifdef HAVE___BUILTIN_POPCOUNT
409 #define _mesa_bitcount(i) __builtin_popcount(i)
412 _mesa_bitcount(unsigned int n
);
415 #ifdef HAVE___BUILTIN_POPCOUNTLL
416 #define _mesa_bitcount_64(i) __builtin_popcountll(i)
419 _mesa_bitcount_64(uint64_t n
);
423 * Find the last (most significant) bit set in a word.
425 * Essentially ffs() in the reverse direction.
427 static inline unsigned int
428 _mesa_fls(unsigned int n
)
430 #ifdef HAVE___BUILTIN_CLZ
431 return n
== 0 ? 0 : 32 - __builtin_clz(n
);
446 _mesa_round_to_even(float val
);
449 _mesa_float_to_half(float f
);
452 _mesa_half_to_float(GLhalfARB h
);
455 _mesa_half_is_negative(GLhalfARB h
)
461 _mesa_strdup( const char *s
);
464 _mesa_str_checksum(const char *str
);
467 _mesa_snprintf( char *str
, size_t size
, const char *fmt
, ... ) PRINTFLIKE(3, 4);
470 _mesa_vsnprintf(char *str
, size_t size
, const char *fmt
, va_list arg
);
473 #if defined(_MSC_VER) && !defined(snprintf)
474 #define snprintf _snprintf
483 #endif /* IMPORTS_H */