Merge branch '7.8' into master
[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 *** LOG2: Log base 2 of float
121 ***/
122 #ifdef USE_IEEE
123 #if 0
124 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
125 * Based on code from http://www.stereopsis.com/log2.html
126 */
127 static INLINE GLfloat LOG2(GLfloat x)
128 {
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 */
134 }
135 #endif
136 /* Pretty fast, and accurate.
137 * Based on code from http://www.flipcode.com/totd/
138 */
139 static INLINE GLfloat LOG2(GLfloat val)
140 {
141 fi_type num;
142 GLint log_2;
143 num.f = val;
144 log_2 = ((num.i >> 23) & 255) - 128;
145 num.i &= ~(255 << 23);
146 num.i += 127 << 23;
147 num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3;
148 return num.f + log_2;
149 }
150 #else
151 /*
152 * NOTE: log_base_2(x) = log(x) / log(2)
153 * NOTE: 1.442695 = 1/log(2).
154 */
155 #define LOG2(x) ((GLfloat) (log(x) * 1.442695F))
156 #endif
157
158
159 /***
160 *** IS_INF_OR_NAN: test if float is infinite or NaN
161 ***/
162 #ifdef USE_IEEE
163 static INLINE int IS_INF_OR_NAN( float x )
164 {
165 fi_type tmp;
166 tmp.f = x;
167 return !(int)((unsigned int)((tmp.i & 0x7fffffff)-0x7f800000) >> 31);
168 }
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))
173 #elif defined(__VMS)
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))
177 #else
178 #define IS_INF_OR_NAN(x) (!finite(x))
179 #endif
180
181
182 /***
183 *** IS_NEGATIVE: test if float is negative
184 ***/
185 #if defined(USE_IEEE)
186 static INLINE int GET_FLOAT_BITS( float x )
187 {
188 fi_type fi;
189 fi.f = x;
190 return fi.i;
191 }
192 #define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) < 0)
193 #else
194 #define IS_NEGATIVE(x) (x < 0.0F)
195 #endif
196
197
198 /***
199 *** DIFFERENT_SIGNS: test if two floats have opposite signs
200 ***/
201 #if defined(USE_IEEE)
202 #define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
203 #else
204 /* Could just use (x*y<0) except for the flatshading requirements.
205 * Maybe there's a better way?
206 */
207 #define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
208 #endif
209
210
211 /***
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
219 ***/
220 #if defined(__gnu_linux__)
221 /* C99 functions */
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)
229 #else
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))
237 #endif
238
239
240 /***
241 *** IROUND: return (as an integer) float rounded to nearest integer
242 ***/
243 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
244 static INLINE int iround(float f)
245 {
246 int r;
247 __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st");
248 return r;
249 }
250 #define IROUND(x) iround(x)
251 #elif defined(USE_X86_ASM) && defined(_MSC_VER)
252 static INLINE int iround(float f)
253 {
254 int r;
255 _asm {
256 fld f
257 fistp r
258 }
259 return r;
260 }
261 #define IROUND(x) iround(x)
262 #elif defined(__WATCOMC__) && defined(__386__)
263 long iround(float f);
264 #pragma aux iround = \
265 "push eax" \
266 "fistp dword ptr [esp]" \
267 "pop eax" \
268 parm [8087] \
269 value [eax] \
270 modify exact [eax];
271 #define IROUND(x) iround(x)
272 #else
273 #define IROUND(f) ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
274 #endif
275
276 #define IROUND64(f) ((GLint64) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
277
278 /***
279 *** IROUND_POS: return (as an integer) positive float rounded to nearest int
280 ***/
281 #ifdef DEBUG
282 #define IROUND_POS(f) (assert((f) >= 0.0F), IROUND(f))
283 #else
284 #define IROUND_POS(f) (IROUND(f))
285 #endif
286
287
288 /***
289 *** IFLOOR: return (as an integer) floor of float
290 ***/
291 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
292 /*
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
298 */
299 static INLINE int ifloor(float f)
300 {
301 int ai, bi;
302 double af, bf;
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;
309 }
310 #define IFLOOR(x) ifloor(x)
311 #elif defined(USE_IEEE)
312 static INLINE int ifloor(float f)
313 {
314 int ai, bi;
315 double af, bf;
316 fi_type u;
317
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;
323 }
324 #define IFLOOR(x) ifloor(x)
325 #else
326 static INLINE int ifloor(float f)
327 {
328 int i = IROUND(f);
329 return (i > f) ? i - 1 : i;
330 }
331 #define IFLOOR(x) ifloor(x)
332 #endif
333
334
335 /***
336 *** ICEIL: return (as an integer) ceiling of float
337 ***/
338 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
339 /*
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
345 */
346 static INLINE int iceil(float f)
347 {
348 int ai, bi;
349 double af, bf;
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;
356 }
357 #define ICEIL(x) iceil(x)
358 #elif defined(USE_IEEE)
359 static INLINE int iceil(float f)
360 {
361 int ai, bi;
362 double af, bf;
363 fi_type u;
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;
369 }
370 #define ICEIL(x) iceil(x)
371 #else
372 static INLINE int iceil(float f)
373 {
374 int i = IROUND(f);
375 return (i < f) ? i + 1 : i;
376 }
377 #define ICEIL(x) iceil(x)
378 #endif
379
380
381 /**
382 * Is x a power of two?
383 */
384 static INLINE int
385 _mesa_is_pow_two(int x)
386 {
387 return !(x & (x - 1));
388 }
389
390 /**
391 * Round given integer to next higer power of two
392 * If X is zero result is undefined.
393 *
394 * Source for the fallback implementation is
395 * Sean Eron Anderson's webpage "Bit Twiddling Hacks"
396 * http://graphics.stanford.edu/~seander/bithacks.html
397 *
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.
403 */
404 static INLINE int32_t
405 _mesa_next_pow_two_32(uint32_t x)
406 {
407 #ifdef __GNUC__
408 uint32_t y = (x != 1);
409 return (1 + y) << ((__builtin_clz(x - y) ^ 31) );
410 #else
411 x--;
412 x |= x >> 1;
413 x |= x >> 2;
414 x |= x >> 4;
415 x |= x >> 8;
416 x |= x >> 16;
417 x++;
418 return x;
419 #endif
420 }
421
422 static INLINE int64_t
423 _mesa_next_pow_two_64(uint64_t x)
424 {
425 #ifdef __GNUC__
426 uint64_t y = (x != 1);
427 if (sizeof(x) == sizeof(long))
428 return (1 + y) << ((__builtin_clzl(x - y) ^ 63));
429 else
430 return (1 + y) << ((__builtin_clzll(x - y) ^ 63));
431 #else
432 x--;
433 x |= x >> 1;
434 x |= x >> 2;
435 x |= x >> 4;
436 x |= x >> 8;
437 x |= x >> 16;
438 x |= x >> 32;
439 x++;
440 return x;
441 #endif
442 }
443
444
445 /***
446 *** UNCLAMPED_FLOAT_TO_UBYTE: clamp float to [0,1] and map to ubyte in [0,255]
447 *** CLAMPED_FLOAT_TO_UBYTE: map float known to be in [0,1] to ubyte in [0,255]
448 ***/
449 #if defined(USE_IEEE) && !defined(DEBUG)
450 #define IEEE_0996 0x3f7f0000 /* 0.996 or so */
451 /* This function/macro is sensitive to precision. Test very carefully
452 * if you change it!
453 */
454 #define UNCLAMPED_FLOAT_TO_UBYTE(UB, F) \
455 do { \
456 fi_type __tmp; \
457 __tmp.f = (F); \
458 if (__tmp.i < 0) \
459 UB = (GLubyte) 0; \
460 else if (__tmp.i >= IEEE_0996) \
461 UB = (GLubyte) 255; \
462 else { \
463 __tmp.f = __tmp.f * (255.0F/256.0F) + 32768.0F; \
464 UB = (GLubyte) __tmp.i; \
465 } \
466 } while (0)
467 #define CLAMPED_FLOAT_TO_UBYTE(UB, F) \
468 do { \
469 fi_type __tmp; \
470 __tmp.f = (F) * (255.0F/256.0F) + 32768.0F; \
471 UB = (GLubyte) __tmp.i; \
472 } while (0)
473 #else
474 #define UNCLAMPED_FLOAT_TO_UBYTE(ub, f) \
475 ub = ((GLubyte) IROUND(CLAMP((f), 0.0F, 1.0F) * 255.0F))
476 #define CLAMPED_FLOAT_TO_UBYTE(ub, f) \
477 ub = ((GLubyte) IROUND((f) * 255.0F))
478 #endif
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_sin(double a);
524
525 extern float
526 _mesa_sinf(float a);
527
528 extern double
529 _mesa_cos(double a);
530
531 extern float
532 _mesa_asinf(float x);
533
534 extern float
535 _mesa_atanf(float x);
536
537 extern double
538 _mesa_sqrtd(double x);
539
540 extern float
541 _mesa_sqrtf(float x);
542
543 extern float
544 _mesa_inv_sqrtf(float x);
545
546 extern void
547 _mesa_init_sqrt_table(void);
548
549 extern double
550 _mesa_pow(double x, double y);
551
552 extern int
553 _mesa_ffs(int32_t i);
554
555 extern int
556 _mesa_ffsll(int64_t i);
557
558 extern unsigned int
559 _mesa_bitcount(unsigned int n);
560
561 extern GLhalfARB
562 _mesa_float_to_half(float f);
563
564 extern float
565 _mesa_half_to_float(GLhalfARB h);
566
567
568 extern void *
569 _mesa_bsearch( const void *key, const void *base, size_t nmemb, size_t size,
570 int (*compar)(const void *, const void *) );
571
572 extern char *
573 _mesa_getenv( const char *var );
574
575 extern char *
576 _mesa_strdup( const char *s );
577
578 extern float
579 _mesa_strtof( const char *s, char **end );
580
581 extern unsigned int
582 _mesa_str_checksum(const char *str);
583
584 extern int
585 _mesa_snprintf( char *str, size_t size, const char *fmt, ... );
586
587 extern void
588 _mesa_warning( __GLcontext *gc, const char *fmtString, ... );
589
590 extern void
591 _mesa_problem( const __GLcontext *ctx, const char *fmtString, ... );
592
593 extern void
594 _mesa_error( __GLcontext *ctx, GLenum error, const char *fmtString, ... );
595
596 extern void
597 _mesa_debug( const __GLcontext *ctx, const char *fmtString, ... );
598
599 #ifdef __cplusplus
600 }
601 #endif
602
603
604 #endif /* IMPORTS_H */