6c2ef52d35b3a5e259ba02b520b15045b08d4a69
[mesa.git] / src / mesa / main / imports.h
1 /*
2 * Mesa 3-D graphics library
3 * Version: 7.5
4 *
5 * Copyright (C) 1999-2008 Brian Paul All Rights Reserved.
6 *
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:
13 *
14 * The above copyright notice and this permission notice shall be included
15 * in all copies or substantial portions of the Software.
16 *
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.
23 */
24
25
26 /**
27 * \file imports.h
28 * Standard C library function wrappers.
29 *
30 * This file provides wrappers for all the standard C library functions
31 * like malloc(), free(), printf(), getenv(), etc.
32 */
33
34
35 #ifndef IMPORTS_H
36 #define IMPORTS_H
37
38
39 #include "compiler.h"
40 #include "glheader.h"
41
42
43 #ifdef __cplusplus
44 extern "C" {
45 #endif
46
47
48 /**********************************************************************/
49 /** Memory macros */
50 /*@{*/
51
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))
60 /** Free memory */
61 #define FREE(PTR) free(PTR)
62
63 /*@}*/
64
65
66 /*
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
70 * this macro.
71 * Both pointers/offsets are expressed in bytes.
72 */
73 #define ADD_POINTERS(A, B) ( (GLubyte *) (A) + (uintptr_t) (B) )
74
75
76 /**
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.
83 */
84 typedef union { GLfloat f; GLint i; } fi_type;
85
86
87
88 /**********************************************************************
89 * Math macros
90 */
91
92 #define MAX_GLUSHORT 0xffff
93 #define MAX_GLUINT 0xffffffff
94
95 /* Degrees to radians conversion: */
96 #define DEG2RAD (M_PI/180.0)
97
98
99 /***
100 *** SQRTF: single-precision square root
101 ***/
102 #if 0 /* _mesa_sqrtf() not accurate enough - temporarily disabled */
103 # define SQRTF(X) _mesa_sqrtf(X)
104 #else
105 # define SQRTF(X) (float) sqrt((float) (X))
106 #endif
107
108
109 /***
110 *** INV_SQRTF: single-precision inverse square root
111 ***/
112 #if 0
113 #define INV_SQRTF(X) _mesa_inv_sqrt(X)
114 #else
115 #define INV_SQRTF(X) (1.0F / SQRTF(X)) /* this is faster on a P4 */
116 #endif
117
118
119 /**
120 * \name Work-arounds for platforms that lack C99 math functions
121 */
122 /*@{*/
123 #if (!defined(_XOPEN_SOURCE) || (_XOPEN_SOURCE < 600)) && !defined(_ISOC99_SOURCE) \
124 && (!defined(__STDC_VERSION__) || (__STDC_VERSION__ < 199901L)) \
125 && (!defined(_MSC_VER) || (_MSC_VER < 1400))
126 #define acosf(f) ((float) acos(f))
127 #define asinf(f) ((float) asin(f))
128 #define atan2f(x,y) ((float) atan2(x,y))
129 #define atanf(f) ((float) atan(f))
130 #define cielf(f) ((float) ciel(f))
131 #define cosf(f) ((float) cos(f))
132 #define coshf(f) ((float) cosh(f))
133 #define expf(f) ((float) exp(f))
134 #define exp2f(f) ((float) exp2(f))
135 #define floorf(f) ((float) floor(f))
136 #define logf(f) ((float) log(f))
137 #define log2f(f) ((float) log2(f))
138 #define powf(x,y) ((float) pow(x,y))
139 #define sinf(f) ((float) sin(f))
140 #define sinhf(f) ((float) sinh(f))
141 #define sqrtf(f) ((float) sqrt(f))
142 #define tanf(f) ((float) tan(f))
143 #define tanhf(f) ((float) tanh(f))
144 #endif
145
146 #if defined(_MSC_VER)
147 static INLINE float truncf(float x) { return x < 0.0f ? ceilf(x) : floorf(x); }
148 static INLINE float exp2f(float x) { return powf(2.0f, x); }
149 static INLINE float log2f(float x) { return logf(x) * 1.442695041f; }
150 #endif
151 /*@}*/
152
153 /***
154 *** LOG2: Log base 2 of float
155 ***/
156 #ifdef USE_IEEE
157 #if 0
158 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
159 * Based on code from http://www.stereopsis.com/log2.html
160 */
161 static INLINE GLfloat LOG2(GLfloat x)
162 {
163 const GLfloat y = x * x * x * x;
164 const GLuint ix = *((GLuint *) &y);
165 const GLuint exp = (ix >> 23) & 0xFF;
166 const GLint log2 = ((GLint) exp) - 127;
167 return (GLfloat) log2 * (1.0 / 4.0); /* 4, because of x^4 above */
168 }
169 #endif
170 /* Pretty fast, and accurate.
171 * Based on code from http://www.flipcode.com/totd/
172 */
173 static INLINE GLfloat LOG2(GLfloat val)
174 {
175 fi_type num;
176 GLint log_2;
177 num.f = val;
178 log_2 = ((num.i >> 23) & 255) - 128;
179 num.i &= ~(255 << 23);
180 num.i += 127 << 23;
181 num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3;
182 return num.f + log_2;
183 }
184 #else
185 /*
186 * NOTE: log_base_2(x) = log(x) / log(2)
187 * NOTE: 1.442695 = 1/log(2).
188 */
189 #define LOG2(x) ((GLfloat) (log(x) * 1.442695F))
190 #endif
191
192
193 /***
194 *** IS_INF_OR_NAN: test if float is infinite or NaN
195 ***/
196 #ifdef USE_IEEE
197 static INLINE int IS_INF_OR_NAN( float x )
198 {
199 fi_type tmp;
200 tmp.f = x;
201 return !(int)((unsigned int)((tmp.i & 0x7fffffff)-0x7f800000) >> 31);
202 }
203 #elif defined(isfinite)
204 #define IS_INF_OR_NAN(x) (!isfinite(x))
205 #elif defined(finite)
206 #define IS_INF_OR_NAN(x) (!finite(x))
207 #elif defined(__VMS)
208 #define IS_INF_OR_NAN(x) (!finite(x))
209 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
210 #define IS_INF_OR_NAN(x) (!isfinite(x))
211 #else
212 #define IS_INF_OR_NAN(x) (!finite(x))
213 #endif
214
215
216 /***
217 *** IS_NEGATIVE: test if float is negative
218 ***/
219 #if defined(USE_IEEE)
220 static INLINE int GET_FLOAT_BITS( float x )
221 {
222 fi_type fi;
223 fi.f = x;
224 return fi.i;
225 }
226 #define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) < 0)
227 #else
228 #define IS_NEGATIVE(x) (x < 0.0F)
229 #endif
230
231
232 /***
233 *** DIFFERENT_SIGNS: test if two floats have opposite signs
234 ***/
235 #if defined(USE_IEEE)
236 #define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
237 #else
238 /* Could just use (x*y<0) except for the flatshading requirements.
239 * Maybe there's a better way?
240 */
241 #define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
242 #endif
243
244
245 /***
246 *** CEILF: ceiling of float
247 *** FLOORF: floor of float
248 *** FABSF: absolute value of float
249 *** LOGF: the natural logarithm (base e) of the value
250 *** EXPF: raise e to the value
251 *** LDEXPF: multiply value by an integral power of two
252 *** FREXPF: extract mantissa and exponent from value
253 ***/
254 #if defined(__gnu_linux__)
255 /* C99 functions */
256 #define CEILF(x) ceilf(x)
257 #define FLOORF(x) floorf(x)
258 #define FABSF(x) fabsf(x)
259 #define LOGF(x) logf(x)
260 #define EXPF(x) expf(x)
261 #define LDEXPF(x,y) ldexpf(x,y)
262 #define FREXPF(x,y) frexpf(x,y)
263 #else
264 #define CEILF(x) ((GLfloat) ceil(x))
265 #define FLOORF(x) ((GLfloat) floor(x))
266 #define FABSF(x) ((GLfloat) fabs(x))
267 #define LOGF(x) ((GLfloat) log(x))
268 #define EXPF(x) ((GLfloat) exp(x))
269 #define LDEXPF(x,y) ((GLfloat) ldexp(x,y))
270 #define FREXPF(x,y) ((GLfloat) frexp(x,y))
271 #endif
272
273
274 /***
275 *** IROUND: return (as an integer) float rounded to nearest integer
276 ***/
277 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
278 static INLINE int iround(float f)
279 {
280 int r;
281 __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st");
282 return r;
283 }
284 #define IROUND(x) iround(x)
285 #elif defined(USE_X86_ASM) && defined(_MSC_VER)
286 static INLINE int iround(float f)
287 {
288 int r;
289 _asm {
290 fld f
291 fistp r
292 }
293 return r;
294 }
295 #define IROUND(x) iround(x)
296 #elif defined(__WATCOMC__) && defined(__386__)
297 long iround(float f);
298 #pragma aux iround = \
299 "push eax" \
300 "fistp dword ptr [esp]" \
301 "pop eax" \
302 parm [8087] \
303 value [eax] \
304 modify exact [eax];
305 #define IROUND(x) iround(x)
306 #else
307 #define IROUND(f) ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
308 #endif
309
310 #define IROUND64(f) ((GLint64) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
311
312 /***
313 *** IROUND_POS: return (as an integer) positive float rounded to nearest int
314 ***/
315 #ifdef DEBUG
316 #define IROUND_POS(f) (assert((f) >= 0.0F), IROUND(f))
317 #else
318 #define IROUND_POS(f) (IROUND(f))
319 #endif
320
321
322 /***
323 *** IFLOOR: return (as an integer) floor of float
324 ***/
325 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
326 /*
327 * IEEE floor for computers that round to nearest or even.
328 * 'f' must be between -4194304 and 4194303.
329 * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
330 * but uses some IEEE specific tricks for better speed.
331 * Contributed by Josh Vanderhoof
332 */
333 static INLINE int ifloor(float f)
334 {
335 int ai, bi;
336 double af, bf;
337 af = (3 << 22) + 0.5 + (double)f;
338 bf = (3 << 22) + 0.5 - (double)f;
339 /* GCC generates an extra fstp/fld without this. */
340 __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
341 __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
342 return (ai - bi) >> 1;
343 }
344 #define IFLOOR(x) ifloor(x)
345 #elif defined(USE_IEEE)
346 static INLINE int ifloor(float f)
347 {
348 int ai, bi;
349 double af, bf;
350 fi_type u;
351
352 af = (3 << 22) + 0.5 + (double)f;
353 bf = (3 << 22) + 0.5 - (double)f;
354 u.f = (float) af; ai = u.i;
355 u.f = (float) bf; bi = u.i;
356 return (ai - bi) >> 1;
357 }
358 #define IFLOOR(x) ifloor(x)
359 #else
360 static INLINE int ifloor(float f)
361 {
362 int i = IROUND(f);
363 return (i > f) ? i - 1 : i;
364 }
365 #define IFLOOR(x) ifloor(x)
366 #endif
367
368
369 /***
370 *** ICEIL: return (as an integer) ceiling of float
371 ***/
372 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
373 /*
374 * IEEE ceil for computers that round to nearest or even.
375 * 'f' must be between -4194304 and 4194303.
376 * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
377 * but uses some IEEE specific tricks for better speed.
378 * Contributed by Josh Vanderhoof
379 */
380 static INLINE int iceil(float f)
381 {
382 int ai, bi;
383 double af, bf;
384 af = (3 << 22) + 0.5 + (double)f;
385 bf = (3 << 22) + 0.5 - (double)f;
386 /* GCC generates an extra fstp/fld without this. */
387 __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
388 __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
389 return (ai - bi + 1) >> 1;
390 }
391 #define ICEIL(x) iceil(x)
392 #elif defined(USE_IEEE)
393 static INLINE int iceil(float f)
394 {
395 int ai, bi;
396 double af, bf;
397 fi_type u;
398 af = (3 << 22) + 0.5 + (double)f;
399 bf = (3 << 22) + 0.5 - (double)f;
400 u.f = (float) af; ai = u.i;
401 u.f = (float) bf; bi = u.i;
402 return (ai - bi + 1) >> 1;
403 }
404 #define ICEIL(x) iceil(x)
405 #else
406 static INLINE int iceil(float f)
407 {
408 int i = IROUND(f);
409 return (i < f) ? i + 1 : i;
410 }
411 #define ICEIL(x) iceil(x)
412 #endif
413
414
415 /**
416 * Is x a power of two?
417 */
418 static INLINE int
419 _mesa_is_pow_two(int x)
420 {
421 return !(x & (x - 1));
422 }
423
424 /**
425 * Round given integer to next higer power of two
426 * If X is zero result is undefined.
427 *
428 * Source for the fallback implementation is
429 * Sean Eron Anderson's webpage "Bit Twiddling Hacks"
430 * http://graphics.stanford.edu/~seander/bithacks.html
431 *
432 * When using builtin function have to do some work
433 * for case when passed values 1 to prevent hiting
434 * undefined result from __builtin_clz. Undefined
435 * results would be different depending on optimization
436 * level used for build.
437 */
438 static INLINE int32_t
439 _mesa_next_pow_two_32(uint32_t x)
440 {
441 #if defined(__GNUC__) && \
442 ((__GNUC__ == 3 && __GNUC_MINOR__ >= 4) || __GNUC__ >= 4)
443 uint32_t y = (x != 1);
444 return (1 + y) << ((__builtin_clz(x - y) ^ 31) );
445 #else
446 x--;
447 x |= x >> 1;
448 x |= x >> 2;
449 x |= x >> 4;
450 x |= x >> 8;
451 x |= x >> 16;
452 x++;
453 return x;
454 #endif
455 }
456
457 static INLINE int64_t
458 _mesa_next_pow_two_64(uint64_t x)
459 {
460 #if defined(__GNUC__) && \
461 ((__GNUC__ == 3 && __GNUC_MINOR__ >= 4) || __GNUC__ >= 4)
462 uint64_t y = (x != 1);
463 if (sizeof(x) == sizeof(long))
464 return (1 + y) << ((__builtin_clzl(x - y) ^ 63));
465 else
466 return (1 + y) << ((__builtin_clzll(x - y) ^ 63));
467 #else
468 x--;
469 x |= x >> 1;
470 x |= x >> 2;
471 x |= x >> 4;
472 x |= x >> 8;
473 x |= x >> 16;
474 x |= x >> 32;
475 x++;
476 return x;
477 #endif
478 }
479
480
481 /**
482 * Return 1 if this is a little endian machine, 0 if big endian.
483 */
484 static INLINE GLboolean
485 _mesa_little_endian(void)
486 {
487 const GLuint ui = 1; /* intentionally not static */
488 return *((const GLubyte *) &ui);
489 }
490
491
492
493 /**********************************************************************
494 * Functions
495 */
496
497 extern void *
498 _mesa_align_malloc( size_t bytes, unsigned long alignment );
499
500 extern void *
501 _mesa_align_calloc( size_t bytes, unsigned long alignment );
502
503 extern void
504 _mesa_align_free( void *ptr );
505
506 extern void *
507 _mesa_align_realloc(void *oldBuffer, size_t oldSize, size_t newSize,
508 unsigned long alignment);
509
510 extern void *
511 _mesa_exec_malloc( GLuint size );
512
513 extern void
514 _mesa_exec_free( void *addr );
515
516 extern void *
517 _mesa_realloc( void *oldBuffer, size_t oldSize, size_t newSize );
518
519 extern void
520 _mesa_memset16( unsigned short *dst, unsigned short val, size_t n );
521
522 extern double
523 _mesa_sqrtd(double x);
524
525 extern float
526 _mesa_sqrtf(float x);
527
528 extern float
529 _mesa_inv_sqrtf(float x);
530
531 extern void
532 _mesa_init_sqrt_table(void);
533
534 extern int
535 _mesa_ffs(int32_t i);
536
537 extern int
538 _mesa_ffsll(int64_t i);
539
540 extern unsigned int
541 _mesa_bitcount(unsigned int n);
542
543 extern GLhalfARB
544 _mesa_float_to_half(float f);
545
546 extern float
547 _mesa_half_to_float(GLhalfARB h);
548
549
550 extern void *
551 _mesa_bsearch( const void *key, const void *base, size_t nmemb, size_t size,
552 int (*compar)(const void *, const void *) );
553
554 extern char *
555 _mesa_getenv( const char *var );
556
557 extern char *
558 _mesa_strdup( const char *s );
559
560 extern float
561 _mesa_strtof( const char *s, char **end );
562
563 extern unsigned int
564 _mesa_str_checksum(const char *str);
565
566 extern int
567 _mesa_snprintf( char *str, size_t size, const char *fmt, ... );
568
569 extern void
570 _mesa_warning( __GLcontext *gc, const char *fmtString, ... );
571
572 extern void
573 _mesa_problem( const __GLcontext *ctx, const char *fmtString, ... );
574
575 extern void
576 _mesa_error( __GLcontext *ctx, GLenum error, const char *fmtString, ... );
577
578 extern void
579 _mesa_debug( const __GLcontext *ctx, const char *fmtString, ... );
580
581
582 #if defined(_MSC_VER) && !defined(snprintf)
583 #define snprintf _snprintf
584 #endif
585
586
587 #ifdef __cplusplus
588 }
589 #endif
590
591
592 #endif /* IMPORTS_H */