imports.h revision 492b69f3be3e355064c67bc6f4a30d40e997ce9d
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#include "errors.h"
42
43#ifdef __cplusplus
44extern "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 */
84typedef union { GLfloat f; GLint i; GLuint u; } 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#define SQRTF(X)  (float) sqrt((float) (X))
103
104
105/***
106 *** INV_SQRTF: single-precision inverse square root
107 ***/
108#define INV_SQRTF(X) (1.0F / SQRTF(X))
109
110
111/**
112 * \name Work-arounds for platforms that lack C99 math functions
113 */
114/*@{*/
115#if (!defined(_XOPEN_SOURCE) || (_XOPEN_SOURCE < 600)) && !defined(_ISOC99_SOURCE) \
116   && (!defined(__STDC_VERSION__) || (__STDC_VERSION__ < 199901L)) \
117   && (!defined(_MSC_VER) || (_MSC_VER < 1400))
118#define acosf(f) ((float) acos(f))
119#define asinf(f) ((float) asin(f))
120#define atan2f(x,y) ((float) atan2(x,y))
121#define atanf(f) ((float) atan(f))
122#define ceilf(f) ((float) ceil(f))
123#define cosf(f) ((float) cos(f))
124#define coshf(f) ((float) cosh(f))
125#define expf(f) ((float) exp(f))
126#define exp2f(f) ((float) exp2(f))
127#define floorf(f) ((float) floor(f))
128#define logf(f) ((float) log(f))
129
130#ifdef ANDROID
131#define log2f(f) (logf(f) * (float) (1.0 / M_LN2))
132#else
133#define log2f(f) ((float) log2(f))
134#endif
135
136#define powf(x,y) ((float) pow(x,y))
137#define sinf(f) ((float) sin(f))
138#define sinhf(f) ((float) sinh(f))
139#define sqrtf(f) ((float) sqrt(f))
140#define tanf(f) ((float) tan(f))
141#define tanhf(f) ((float) tanh(f))
142#define acoshf(f) ((float) acosh(f))
143#define asinhf(f) ((float) asinh(f))
144#define atanhf(f) ((float) atanh(f))
145#endif
146
147#if defined(_MSC_VER)
148static inline float truncf(float x) { return x < 0.0f ? ceilf(x) : floorf(x); }
149static inline float exp2f(float x) { return powf(2.0f, x); }
150static inline float log2f(float x) { return logf(x) * 1.442695041f; }
151static inline float asinhf(float x) { return logf(x + sqrtf(x * x + 1.0f)); }
152static inline float acoshf(float x) { return logf(x + sqrtf(x * x - 1.0f)); }
153static inline float atanhf(float x) { return (logf(1.0f + x) - logf(1.0f - x)) / 2.0f; }
154static inline int isblank(int ch) { return ch == ' ' || ch == '\t'; }
155#define strtoll(p, e, b) _strtoi64(p, e, b)
156#endif
157/*@}*/
158
159/***
160 *** LOG2: Log base 2 of float
161 ***/
162#ifdef USE_IEEE
163#if 0
164/* This is pretty fast, but not accurate enough (only 2 fractional bits).
165 * Based on code from http://www.stereopsis.com/log2.html
166 */
167static inline GLfloat LOG2(GLfloat x)
168{
169   const GLfloat y = x * x * x * x;
170   const GLuint ix = *((GLuint *) &y);
171   const GLuint exp = (ix >> 23) & 0xFF;
172   const GLint log2 = ((GLint) exp) - 127;
173   return (GLfloat) log2 * (1.0 / 4.0);  /* 4, because of x^4 above */
174}
175#endif
176/* Pretty fast, and accurate.
177 * Based on code from http://www.flipcode.com/totd/
178 */
179static inline GLfloat LOG2(GLfloat val)
180{
181   fi_type num;
182   GLint log_2;
183   num.f = val;
184   log_2 = ((num.i >> 23) & 255) - 128;
185   num.i &= ~(255 << 23);
186   num.i += 127 << 23;
187   num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3;
188   return num.f + log_2;
189}
190#else
191/*
192 * NOTE: log_base_2(x) = log(x) / log(2)
193 * NOTE: 1.442695 = 1/log(2).
194 */
195#define LOG2(x)  ((GLfloat) (log(x) * 1.442695F))
196#endif
197
198
199/***
200 *** IS_INF_OR_NAN: test if float is infinite or NaN
201 ***/
202#ifdef USE_IEEE
203static inline int IS_INF_OR_NAN( float x )
204{
205   fi_type tmp;
206   tmp.f = x;
207   return !(int)((unsigned int)((tmp.i & 0x7fffffff)-0x7f800000) >> 31);
208}
209#elif defined(isfinite)
210#define IS_INF_OR_NAN(x)        (!isfinite(x))
211#elif defined(finite)
212#define IS_INF_OR_NAN(x)        (!finite(x))
213#elif defined(__VMS)
214#define IS_INF_OR_NAN(x)        (!finite(x))
215#elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
216#define IS_INF_OR_NAN(x)        (!isfinite(x))
217#else
218#define IS_INF_OR_NAN(x)        (!finite(x))
219#endif
220
221
222/***
223 *** IS_NEGATIVE: test if float is negative
224 ***/
225#if defined(USE_IEEE)
226static inline int GET_FLOAT_BITS( float x )
227{
228   fi_type fi;
229   fi.f = x;
230   return fi.i;
231}
232#define IS_NEGATIVE(x) (GET_FLOAT_BITS(x) < 0)
233#else
234#define IS_NEGATIVE(x) (x < 0.0F)
235#endif
236
237
238/***
239 *** DIFFERENT_SIGNS: test if two floats have opposite signs
240 ***/
241#if defined(USE_IEEE)
242#define DIFFERENT_SIGNS(x,y) ((GET_FLOAT_BITS(x) ^ GET_FLOAT_BITS(y)) & (1<<31))
243#else
244/* Could just use (x*y<0) except for the flatshading requirements.
245 * Maybe there's a better way?
246 */
247#define DIFFERENT_SIGNS(x,y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
248#endif
249
250
251/***
252 *** CEILF: ceiling of float
253 *** FLOORF: floor of float
254 *** FABSF: absolute value of float
255 *** LOGF: the natural logarithm (base e) of the value
256 *** EXPF: raise e to the value
257 *** LDEXPF: multiply value by an integral power of two
258 *** FREXPF: extract mantissa and exponent from value
259 ***/
260#if defined(__gnu_linux__)
261/* C99 functions */
262#define CEILF(x)   ceilf(x)
263#define FLOORF(x)  floorf(x)
264#define FABSF(x)   fabsf(x)
265#define LOGF(x)    logf(x)
266#define EXPF(x)    expf(x)
267#define LDEXPF(x,y)  ldexpf(x,y)
268#define FREXPF(x,y)  frexpf(x,y)
269#else
270#define CEILF(x)   ((GLfloat) ceil(x))
271#define FLOORF(x)  ((GLfloat) floor(x))
272#define FABSF(x)   ((GLfloat) fabs(x))
273#define LOGF(x)    ((GLfloat) log(x))
274#define EXPF(x)    ((GLfloat) exp(x))
275#define LDEXPF(x,y)  ((GLfloat) ldexp(x,y))
276#define FREXPF(x,y)  ((GLfloat) frexp(x,y))
277#endif
278
279
280/**
281 * Convert float to int by rounding to nearest integer, away from zero.
282 */
283static inline int IROUND(float f)
284{
285   return (int) ((f >= 0.0F) ? (f + 0.5F) : (f - 0.5F));
286}
287
288
289/**
290 * Convert float to int64 by rounding to nearest integer.
291 */
292static inline GLint64 IROUND64(float f)
293{
294   return (GLint64) ((f >= 0.0F) ? (f + 0.5F) : (f - 0.5F));
295}
296
297
298/**
299 * Convert positive float to int by rounding to nearest integer.
300 */
301static inline int IROUND_POS(float f)
302{
303   assert(f >= 0.0F);
304   return (int) (f + 0.5F);
305}
306
307
308/**
309 * Convert float to int using a fast method.  The rounding mode may vary.
310 * XXX We could use an x86-64/SSE2 version here.
311 */
312#if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
313static inline int F_TO_I(float f)
314{
315   int r;
316   __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st");
317   return r;
318}
319#elif defined(USE_X86_ASM) && defined(_MSC_VER)
320static inline int F_TO_I(float f)
321{
322   int r;
323   _asm {
324	 fld f
325	 fistp r
326	}
327   return r;
328}
329#elif defined(__WATCOMC__) && defined(__386__)
330long F_TO_I(float f);
331#pragma aux iround =                    \
332	"push   eax"                        \
333	"fistp  dword ptr [esp]"            \
334	"pop    eax"                        \
335	parm [8087]                         \
336	value [eax]                         \
337	modify exact [eax];
338#else
339#define F_TO_I(f)  IROUND(f)
340#endif
341
342
343/***
344 *** IFLOOR: return (as an integer) floor of float
345 ***/
346#if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
347/*
348 * IEEE floor for computers that round to nearest or even.
349 * 'f' must be between -4194304 and 4194303.
350 * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
351 * but uses some IEEE specific tricks for better speed.
352 * Contributed by Josh Vanderhoof
353 */
354static inline int ifloor(float f)
355{
356   int ai, bi;
357   double af, bf;
358   af = (3 << 22) + 0.5 + (double)f;
359   bf = (3 << 22) + 0.5 - (double)f;
360   /* GCC generates an extra fstp/fld without this. */
361   __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
362   __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
363   return (ai - bi) >> 1;
364}
365#define IFLOOR(x)  ifloor(x)
366#elif defined(USE_IEEE)
367static inline int ifloor(float f)
368{
369   int ai, bi;
370   double af, bf;
371   fi_type u;
372
373   af = (3 << 22) + 0.5 + (double)f;
374   bf = (3 << 22) + 0.5 - (double)f;
375   u.f = (float) af;  ai = u.i;
376   u.f = (float) bf;  bi = u.i;
377   return (ai - bi) >> 1;
378}
379#define IFLOOR(x)  ifloor(x)
380#else
381static inline int ifloor(float f)
382{
383   int i = IROUND(f);
384   return (i > f) ? i - 1 : i;
385}
386#define IFLOOR(x)  ifloor(x)
387#endif
388
389
390/***
391 *** ICEIL: return (as an integer) ceiling of float
392 ***/
393#if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
394/*
395 * IEEE ceil for computers that round to nearest or even.
396 * 'f' must be between -4194304 and 4194303.
397 * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
398 * but uses some IEEE specific tricks for better speed.
399 * Contributed by Josh Vanderhoof
400 */
401static inline int iceil(float f)
402{
403   int ai, bi;
404   double af, bf;
405   af = (3 << 22) + 0.5 + (double)f;
406   bf = (3 << 22) + 0.5 - (double)f;
407   /* GCC generates an extra fstp/fld without this. */
408   __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
409   __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
410   return (ai - bi + 1) >> 1;
411}
412#define ICEIL(x)  iceil(x)
413#elif defined(USE_IEEE)
414static inline int iceil(float f)
415{
416   int ai, bi;
417   double af, bf;
418   fi_type u;
419   af = (3 << 22) + 0.5 + (double)f;
420   bf = (3 << 22) + 0.5 - (double)f;
421   u.f = (float) af; ai = u.i;
422   u.f = (float) bf; bi = u.i;
423   return (ai - bi + 1) >> 1;
424}
425#define ICEIL(x)  iceil(x)
426#else
427static inline int iceil(float f)
428{
429   int i = IROUND(f);
430   return (i < f) ? i + 1 : i;
431}
432#define ICEIL(x)  iceil(x)
433#endif
434
435
436/**
437 * Is x a power of two?
438 */
439static inline int
440_mesa_is_pow_two(int x)
441{
442   return !(x & (x - 1));
443}
444
445/**
446 * Round given integer to next higer power of two
447 * If X is zero result is undefined.
448 *
449 * Source for the fallback implementation is
450 * Sean Eron Anderson's webpage "Bit Twiddling Hacks"
451 * http://graphics.stanford.edu/~seander/bithacks.html
452 *
453 * When using builtin function have to do some work
454 * for case when passed values 1 to prevent hiting
455 * undefined result from __builtin_clz. Undefined
456 * results would be different depending on optimization
457 * level used for build.
458 */
459static inline int32_t
460_mesa_next_pow_two_32(uint32_t x)
461{
462#if defined(__GNUC__) && \
463	((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
464	uint32_t y = (x != 1);
465	return (1 + y) << ((__builtin_clz(x - y) ^ 31) );
466#else
467	x--;
468	x |= x >> 1;
469	x |= x >> 2;
470	x |= x >> 4;
471	x |= x >> 8;
472	x |= x >> 16;
473	x++;
474	return x;
475#endif
476}
477
478static inline int64_t
479_mesa_next_pow_two_64(uint64_t x)
480{
481#if defined(__GNUC__) && \
482	((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
483	uint64_t y = (x != 1);
484	if (sizeof(x) == sizeof(long))
485		return (1 + y) << ((__builtin_clzl(x - y) ^ 63));
486	else
487		return (1 + y) << ((__builtin_clzll(x - y) ^ 63));
488#else
489	x--;
490	x |= x >> 1;
491	x |= x >> 2;
492	x |= x >> 4;
493	x |= x >> 8;
494	x |= x >> 16;
495	x |= x >> 32;
496	x++;
497	return x;
498#endif
499}
500
501
502/*
503 * Returns the floor form of binary logarithm for a 32-bit integer.
504 */
505static inline GLuint
506_mesa_logbase2(GLuint n)
507{
508#if defined(__GNUC__) && \
509   ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
510   return (31 - __builtin_clz(n | 1));
511#else
512   GLuint pos = 0;
513   if (n >= 1<<16) { n >>= 16; pos += 16; }
514   if (n >= 1<< 8) { n >>=  8; pos +=  8; }
515   if (n >= 1<< 4) { n >>=  4; pos +=  4; }
516   if (n >= 1<< 2) { n >>=  2; pos +=  2; }
517   if (n >= 1<< 1) {           pos +=  1; }
518   return pos;
519#endif
520}
521
522
523/**
524 * Return 1 if this is a little endian machine, 0 if big endian.
525 */
526static inline GLboolean
527_mesa_little_endian(void)
528{
529   const GLuint ui = 1; /* intentionally not static */
530   return *((const GLubyte *) &ui);
531}
532
533
534
535/**********************************************************************
536 * Functions
537 */
538
539extern void *
540_mesa_align_malloc( size_t bytes, unsigned long alignment );
541
542extern void *
543_mesa_align_calloc( size_t bytes, unsigned long alignment );
544
545extern void
546_mesa_align_free( void *ptr );
547
548extern void *
549_mesa_align_realloc(void *oldBuffer, size_t oldSize, size_t newSize,
550                    unsigned long alignment);
551
552extern void *
553_mesa_exec_malloc( GLuint size );
554
555extern void
556_mesa_exec_free( void *addr );
557
558extern void *
559_mesa_realloc( void *oldBuffer, size_t oldSize, size_t newSize );
560
561
562#ifndef FFS_DEFINED
563#define FFS_DEFINED 1
564#ifdef __GNUC__
565
566#if defined(__MINGW32__) || defined(__CYGWIN__) || defined(ANDROID) || defined(__APPLE__)
567#define ffs __builtin_ffs
568#define ffsll __builtin_ffsll
569#endif
570
571#else
572
573extern int ffs(int i);
574extern int ffsll(long long int i);
575
576#endif /*__ GNUC__ */
577#endif /* FFS_DEFINED */
578
579
580#if defined(__GNUC__) && ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
581#define _mesa_bitcount(i) __builtin_popcount(i)
582#define _mesa_bitcount_64(i) __builtin_popcountll(i)
583#else
584extern unsigned int
585_mesa_bitcount(unsigned int n);
586extern unsigned int
587_mesa_bitcount_64(uint64_t n);
588#endif
589
590/**
591 * Find the last (most significant) bit set in a word.
592 *
593 * Essentially ffs() in the reverse direction.
594 */
595static inline unsigned int
596_mesa_fls(unsigned int n)
597{
598#if defined(__GNUC__) && ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304)
599   return n == 0 ? 0 : 32 - __builtin_clz(n);
600#else
601   unsigned int v = 1;
602
603   if (n == 0)
604      return 0;
605
606   while (n >>= 1)
607       v++;
608
609   return v;
610#endif
611}
612
613extern GLhalfARB
614_mesa_float_to_half(float f);
615
616extern float
617_mesa_half_to_float(GLhalfARB h);
618
619
620extern void *
621_mesa_bsearch( const void *key, const void *base, size_t nmemb, size_t size,
622               int (*compar)(const void *, const void *) );
623
624extern char *
625_mesa_getenv( const char *var );
626
627extern char *
628_mesa_strdup( const char *s );
629
630extern float
631_mesa_strtof( const char *s, char **end );
632
633extern unsigned int
634_mesa_str_checksum(const char *str);
635
636extern int
637_mesa_snprintf( char *str, size_t size, const char *fmt, ... ) PRINTFLIKE(3, 4);
638
639extern int
640_mesa_vsnprintf(char *str, size_t size, const char *fmt, va_list arg);
641
642
643#if defined(_MSC_VER) && !defined(snprintf)
644#define snprintf _snprintf
645#endif
646
647
648#ifdef __cplusplus
649}
650#endif
651
652
653#endif /* IMPORTS_H */
654