Merge branch 'gallium-polygon-stipple'
[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 #define acoshf(f) ((float) acosh(f))
145 #define asinhf(f) ((float) asinh(f))
146 #define atanhf(f) ((float) atanh(f))
147 #endif
148
149 #if defined(_MSC_VER)
150 static INLINE float truncf(float x) { return x < 0.0f ? ceilf(x) : floorf(x); }
151 static INLINE float exp2f(float x) { return powf(2.0f, x); }
152 static INLINE float log2f(float x) { return logf(x) * 1.442695041f; }
153 static INLINE float asinhf(float x) { return logf(x + sqrtf(x * x + 1.0f)); }
154 static INLINE float acoshf(float x) { return logf(x + sqrtf(x * x - 1.0f)); }
155 static INLINE float atanhf(float x) { return (logf(1.0f + x) - logf(1.0f - x)) / 2.0f; }
156 static INLINE int isblank(int ch) { return ch == ' ' || ch == '\t'; }
157 #define strtoll(p, e, b) _strtoi64(p, e, b)
158 #endif
159 /*@}*/
160
161 /***
162 *** LOG2: Log base 2 of float
163 ***/
164 #ifdef USE_IEEE
165 #if 0
166 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
167 * Based on code from http://www.stereopsis.com/log2.html
168 */
169 static INLINE GLfloat LOG2(GLfloat x)
170 {
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 */
176 }
177 #endif
178 /* Pretty fast, and accurate.
179 * Based on code from http://www.flipcode.com/totd/
180 */
181 static INLINE GLfloat LOG2(GLfloat val)
182 {
183 fi_type num;
184 GLint log_2;
185 num.f = val;
186 log_2 = ((num.i >> 23) & 255) - 128;
187 num.i &= ~(255 << 23);
188 num.i += 127 << 23;
189 num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3;
190 return num.f + log_2;
191 }
192 #else
193 /*
194 * NOTE: log_base_2(x) = log(x) / log(2)
195 * NOTE: 1.442695 = 1/log(2).
196 */
197 #define LOG2(x) ((GLfloat) (log(x) * 1.442695F))
198 #endif
199
200
201 /***
202 *** IS_INF_OR_NAN: test if float is infinite or NaN
203 ***/
204 #ifdef USE_IEEE
205 static INLINE int IS_INF_OR_NAN( float x )
206 {
207 fi_type tmp;
208 tmp.f = x;
209 return !(int)((unsigned int)((tmp.i & 0x7fffffff)-0x7f800000) >> 31);
210 }
211 #elif defined(isfinite)
212 #define IS_INF_OR_NAN(x) (!isfinite(x))
213 #elif defined(finite)
214 #define IS_INF_OR_NAN(x) (!finite(x))
215 #elif defined(__VMS)
216 #define IS_INF_OR_NAN(x) (!finite(x))
217 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
218 #define IS_INF_OR_NAN(x) (!isfinite(x))
219 #else
220 #define IS_INF_OR_NAN(x) (!finite(x))
221 #endif
222
223
224 /***
225 *** IS_NEGATIVE: test if float is negative
226 ***/
227 #if defined(USE_IEEE)
228 static INLINE int GET_FLOAT_BITS( float x )
229 {
230 fi_type fi;
231 fi.f = x;
232 return fi.i;
233 }
234 #define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) < 0)
235 #else
236 #define IS_NEGATIVE(x) (x < 0.0F)
237 #endif
238
239
240 /***
241 *** DIFFERENT_SIGNS: test if two floats have opposite signs
242 ***/
243 #if defined(USE_IEEE)
244 #define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
245 #else
246 /* Could just use (x*y<0) except for the flatshading requirements.
247 * Maybe there's a better way?
248 */
249 #define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
250 #endif
251
252
253 /***
254 *** CEILF: ceiling of float
255 *** FLOORF: floor of float
256 *** FABSF: absolute value of float
257 *** LOGF: the natural logarithm (base e) of the value
258 *** EXPF: raise e to the value
259 *** LDEXPF: multiply value by an integral power of two
260 *** FREXPF: extract mantissa and exponent from value
261 ***/
262 #if defined(__gnu_linux__)
263 /* C99 functions */
264 #define CEILF(x) ceilf(x)
265 #define FLOORF(x) floorf(x)
266 #define FABSF(x) fabsf(x)
267 #define LOGF(x) logf(x)
268 #define EXPF(x) expf(x)
269 #define LDEXPF(x,y) ldexpf(x,y)
270 #define FREXPF(x,y) frexpf(x,y)
271 #else
272 #define CEILF(x) ((GLfloat) ceil(x))
273 #define FLOORF(x) ((GLfloat) floor(x))
274 #define FABSF(x) ((GLfloat) fabs(x))
275 #define LOGF(x) ((GLfloat) log(x))
276 #define EXPF(x) ((GLfloat) exp(x))
277 #define LDEXPF(x,y) ((GLfloat) ldexp(x,y))
278 #define FREXPF(x,y) ((GLfloat) frexp(x,y))
279 #endif
280
281
282 /***
283 *** IROUND: return (as an integer) float rounded to nearest integer
284 ***/
285 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
286 static INLINE int iround(float f)
287 {
288 int r;
289 __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st");
290 return r;
291 }
292 #define IROUND(x) iround(x)
293 #elif defined(USE_X86_ASM) && defined(_MSC_VER)
294 static INLINE int iround(float f)
295 {
296 int r;
297 _asm {
298 fld f
299 fistp r
300 }
301 return r;
302 }
303 #define IROUND(x) iround(x)
304 #elif defined(__WATCOMC__) && defined(__386__)
305 long iround(float f);
306 #pragma aux iround = \
307 "push eax" \
308 "fistp dword ptr [esp]" \
309 "pop eax" \
310 parm [8087] \
311 value [eax] \
312 modify exact [eax];
313 #define IROUND(x) iround(x)
314 #else
315 #define IROUND(f) ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
316 #endif
317
318 #define IROUND64(f) ((GLint64) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
319
320 /***
321 *** IROUND_POS: return (as an integer) positive float rounded to nearest int
322 ***/
323 #ifdef DEBUG
324 #define IROUND_POS(f) (assert((f) >= 0.0F), IROUND(f))
325 #else
326 #define IROUND_POS(f) (IROUND(f))
327 #endif
328
329
330 /***
331 *** IFLOOR: return (as an integer) floor of float
332 ***/
333 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
334 /*
335 * IEEE floor for computers that round to nearest or even.
336 * 'f' must be between -4194304 and 4194303.
337 * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
338 * but uses some IEEE specific tricks for better speed.
339 * Contributed by Josh Vanderhoof
340 */
341 static INLINE int ifloor(float f)
342 {
343 int ai, bi;
344 double af, bf;
345 af = (3 << 22) + 0.5 + (double)f;
346 bf = (3 << 22) + 0.5 - (double)f;
347 /* GCC generates an extra fstp/fld without this. */
348 __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
349 __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
350 return (ai - bi) >> 1;
351 }
352 #define IFLOOR(x) ifloor(x)
353 #elif defined(USE_IEEE)
354 static INLINE int ifloor(float f)
355 {
356 int ai, bi;
357 double af, bf;
358 fi_type u;
359
360 af = (3 << 22) + 0.5 + (double)f;
361 bf = (3 << 22) + 0.5 - (double)f;
362 u.f = (float) af; ai = u.i;
363 u.f = (float) bf; bi = u.i;
364 return (ai - bi) >> 1;
365 }
366 #define IFLOOR(x) ifloor(x)
367 #else
368 static INLINE int ifloor(float f)
369 {
370 int i = IROUND(f);
371 return (i > f) ? i - 1 : i;
372 }
373 #define IFLOOR(x) ifloor(x)
374 #endif
375
376
377 /***
378 *** ICEIL: return (as an integer) ceiling of float
379 ***/
380 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
381 /*
382 * IEEE ceil for computers that round to nearest or even.
383 * 'f' must be between -4194304 and 4194303.
384 * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
385 * but uses some IEEE specific tricks for better speed.
386 * Contributed by Josh Vanderhoof
387 */
388 static INLINE int iceil(float f)
389 {
390 int ai, bi;
391 double af, bf;
392 af = (3 << 22) + 0.5 + (double)f;
393 bf = (3 << 22) + 0.5 - (double)f;
394 /* GCC generates an extra fstp/fld without this. */
395 __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
396 __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
397 return (ai - bi + 1) >> 1;
398 }
399 #define ICEIL(x) iceil(x)
400 #elif defined(USE_IEEE)
401 static INLINE int iceil(float f)
402 {
403 int ai, bi;
404 double af, bf;
405 fi_type u;
406 af = (3 << 22) + 0.5 + (double)f;
407 bf = (3 << 22) + 0.5 - (double)f;
408 u.f = (float) af; ai = u.i;
409 u.f = (float) bf; bi = u.i;
410 return (ai - bi + 1) >> 1;
411 }
412 #define ICEIL(x) iceil(x)
413 #else
414 static INLINE int iceil(float f)
415 {
416 int i = IROUND(f);
417 return (i < f) ? i + 1 : i;
418 }
419 #define ICEIL(x) iceil(x)
420 #endif
421
422
423 /**
424 * Is x a power of two?
425 */
426 static INLINE int
427 _mesa_is_pow_two(int x)
428 {
429 return !(x & (x - 1));
430 }
431
432 /**
433 * Round given integer to next higer power of two
434 * If X is zero result is undefined.
435 *
436 * Source for the fallback implementation is
437 * Sean Eron Anderson's webpage "Bit Twiddling Hacks"
438 * http://graphics.stanford.edu/~seander/bithacks.html
439 *
440 * When using builtin function have to do some work
441 * for case when passed values 1 to prevent hiting
442 * undefined result from __builtin_clz. Undefined
443 * results would be different depending on optimization
444 * level used for build.
445 */
446 static INLINE int32_t
447 _mesa_next_pow_two_32(uint32_t x)
448 {
449 #if defined(__GNUC__) && \
450 ((__GNUC__ == 3 && __GNUC_MINOR__ >= 4) || __GNUC__ >= 4)
451 uint32_t y = (x != 1);
452 return (1 + y) << ((__builtin_clz(x - y) ^ 31) );
453 #else
454 x--;
455 x |= x >> 1;
456 x |= x >> 2;
457 x |= x >> 4;
458 x |= x >> 8;
459 x |= x >> 16;
460 x++;
461 return x;
462 #endif
463 }
464
465 static INLINE int64_t
466 _mesa_next_pow_two_64(uint64_t x)
467 {
468 #if defined(__GNUC__) && \
469 ((__GNUC__ == 3 && __GNUC_MINOR__ >= 4) || __GNUC__ >= 4)
470 uint64_t y = (x != 1);
471 if (sizeof(x) == sizeof(long))
472 return (1 + y) << ((__builtin_clzl(x - y) ^ 63));
473 else
474 return (1 + y) << ((__builtin_clzll(x - y) ^ 63));
475 #else
476 x--;
477 x |= x >> 1;
478 x |= x >> 2;
479 x |= x >> 4;
480 x |= x >> 8;
481 x |= x >> 16;
482 x |= x >> 32;
483 x++;
484 return x;
485 #endif
486 }
487
488
489 /*
490 * Returns the floor form of binary logarithm for a 32-bit integer.
491 */
492 static INLINE GLuint
493 _mesa_logbase2(GLuint n)
494 {
495 #if defined(__GNUC__) && \
496 ((__GNUC__ == 3 && __GNUC_MINOR__ >= 4) || __GNUC__ >= 4)
497 return (31 - __builtin_clz(n | 1));
498 #else
499 GLuint pos = 0;
500 if (n >= 1<<16) { n >>= 16; pos += 16; }
501 if (n >= 1<< 8) { n >>= 8; pos += 8; }
502 if (n >= 1<< 4) { n >>= 4; pos += 4; }
503 if (n >= 1<< 2) { n >>= 2; pos += 2; }
504 if (n >= 1<< 1) { pos += 1; }
505 return pos;
506 #endif
507 }
508
509
510 /**
511 * Return 1 if this is a little endian machine, 0 if big endian.
512 */
513 static INLINE GLboolean
514 _mesa_little_endian(void)
515 {
516 const GLuint ui = 1; /* intentionally not static */
517 return *((const GLubyte *) &ui);
518 }
519
520
521
522 /**********************************************************************
523 * Functions
524 */
525
526 extern void *
527 _mesa_align_malloc( size_t bytes, unsigned long alignment );
528
529 extern void *
530 _mesa_align_calloc( size_t bytes, unsigned long alignment );
531
532 extern void
533 _mesa_align_free( void *ptr );
534
535 extern void *
536 _mesa_align_realloc(void *oldBuffer, size_t oldSize, size_t newSize,
537 unsigned long alignment);
538
539 extern void *
540 _mesa_exec_malloc( GLuint size );
541
542 extern void
543 _mesa_exec_free( void *addr );
544
545 extern void *
546 _mesa_realloc( void *oldBuffer, size_t oldSize, size_t newSize );
547
548 extern void
549 _mesa_memset16( unsigned short *dst, unsigned short val, size_t n );
550
551 extern double
552 _mesa_sqrtd(double x);
553
554 extern float
555 _mesa_sqrtf(float x);
556
557 extern float
558 _mesa_inv_sqrtf(float x);
559
560 extern void
561 _mesa_init_sqrt_table(void);
562
563 #ifdef __GNUC__
564
565 #ifdef __MINGW32__
566 #define ffs __builtin_ffs
567 #define ffsll __builtin_ffsll
568 #endif
569
570 #define _mesa_ffs(i) ffs(i)
571 #define _mesa_ffsll(i) ffsll(i)
572
573 #if ((_GNUC__ == 3 && __GNUC_MINOR__ >= 4) || __GNUC__ >= 4)
574 #define _mesa_bitcount(i) __builtin_popcount(i)
575 #else
576 extern unsigned int
577 _mesa_bitcount(unsigned int n);
578 #endif
579
580 #else
581 extern int
582 _mesa_ffs(int32_t i);
583
584 extern int
585 _mesa_ffsll(int64_t i);
586
587 extern unsigned int
588 _mesa_bitcount(unsigned int n);
589 #endif
590
591 extern GLhalfARB
592 _mesa_float_to_half(float f);
593
594 extern float
595 _mesa_half_to_float(GLhalfARB h);
596
597
598 extern void *
599 _mesa_bsearch( const void *key, const void *base, size_t nmemb, size_t size,
600 int (*compar)(const void *, const void *) );
601
602 extern char *
603 _mesa_getenv( const char *var );
604
605 extern char *
606 _mesa_strdup( const char *s );
607
608 extern float
609 _mesa_strtof( const char *s, char **end );
610
611 extern unsigned int
612 _mesa_str_checksum(const char *str);
613
614 extern int
615 _mesa_snprintf( char *str, size_t size, const char *fmt, ... ) PRINTFLIKE(3, 4);
616
617 struct gl_context;
618
619 extern void
620 _mesa_warning( struct gl_context *gc, const char *fmtString, ... ) PRINTFLIKE(2, 3);
621
622 extern void
623 _mesa_problem( const struct gl_context *ctx, const char *fmtString, ... ) PRINTFLIKE(2, 3);
624
625 extern void
626 _mesa_error( struct gl_context *ctx, GLenum error, const char *fmtString, ... ) PRINTFLIKE(3, 4);
627
628 extern void
629 _mesa_debug( const struct gl_context *ctx, const char *fmtString, ... ) PRINTFLIKE(2, 3);
630
631
632 #if defined(_MSC_VER) && !defined(snprintf)
633 #define snprintf _snprintf
634 #endif
635
636
637 #ifdef __cplusplus
638 }
639 #endif
640
641
642 #endif /* IMPORTS_H */