mesa: replace FABSF with fabsf
[mesa.git] / src / mesa / main / imports.h
1 /*
2 * Mesa 3-D graphics library
3 *
4 * Copyright (C) 1999-2008 Brian Paul All Rights Reserved.
5 *
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:
12 *
13 * The above copyright notice and this permission notice shall be included
14 * in all copies or substantial portions of the Software.
15 *
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.
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
44 extern "C" {
45 #endif
46
47
48 /**********************************************************************/
49 /** Memory macros */
50 /*@{*/
51
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))
56
57 /*@}*/
58
59
60 /*
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
64 * this macro.
65 * Both pointers/offsets are expressed in bytes.
66 */
67 #define ADD_POINTERS(A, B) ( (GLubyte *) (A) + (uintptr_t) (B) )
68
69
70 /**
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.
77 */
78 typedef union { GLfloat f; GLint i; GLuint u; } fi_type;
79
80
81
82 /**********************************************************************
83 * Math macros
84 */
85
86 #define MAX_GLUSHORT 0xffff
87 #define MAX_GLUINT 0xffffffff
88
89 /* Degrees to radians conversion: */
90 #define DEG2RAD (M_PI/180.0)
91
92
93 /**
94 * \name Work-arounds for platforms that lack C99 math functions
95 */
96 /*@{*/
97 #if (!defined(_XOPEN_SOURCE) || (_XOPEN_SOURCE < 600)) && !defined(_ISOC99_SOURCE) \
98 && (!defined(__STDC_VERSION__) || (__STDC_VERSION__ < 199901L)) \
99 && (!defined(_MSC_VER) || (_MSC_VER < 1400))
100 #define acosf(f) ((float) acos(f))
101 #define asinf(f) ((float) asin(f))
102 #define atan2f(x,y) ((float) atan2(x,y))
103 #define atanf(f) ((float) atan(f))
104 #define ceilf(f) ((float) ceil(f))
105 #define cosf(f) ((float) cos(f))
106 #define coshf(f) ((float) cosh(f))
107 #define expf(f) ((float) exp(f))
108 #define exp2f(f) ((float) exp2(f))
109 #define floorf(f) ((float) floor(f))
110 #define logf(f) ((float) log(f))
111
112 #ifdef ANDROID
113 #define log2f(f) (logf(f) * (float) (1.0 / M_LN2))
114 #else
115 #define log2f(f) ((float) log2(f))
116 #endif
117
118 #define powf(x,y) ((float) pow(x,y))
119 #define sinf(f) ((float) sin(f))
120 #define sinhf(f) ((float) sinh(f))
121 #define sqrtf(f) ((float) sqrt(f))
122 #define tanf(f) ((float) tan(f))
123 #define tanhf(f) ((float) tanh(f))
124 #define acoshf(f) ((float) acosh(f))
125 #define asinhf(f) ((float) asinh(f))
126 #define atanhf(f) ((float) atanh(f))
127 #endif
128
129 #if defined(_MSC_VER)
130 #if _MSC_VER < 1800 /* Not req'd on VS2013 and above */
131 static inline float truncf(float x) { return x < 0.0f ? ceilf(x) : floorf(x); }
132 static inline float exp2f(float x) { return powf(2.0f, x); }
133 static inline float log2f(float x) { return logf(x) * 1.442695041f; }
134 static inline float asinhf(float x) { return logf(x + sqrtf(x * x + 1.0f)); }
135 static inline float acoshf(float x) { return logf(x + sqrtf(x * x - 1.0f)); }
136 static inline float atanhf(float x) { return (logf(1.0f + x) - logf(1.0f - x)) / 2.0f; }
137 static inline int isblank(int ch) { return ch == ' ' || ch == '\t'; }
138 #define strtoll(p, e, b) _strtoi64(p, e, b)
139 #endif /* _MSC_VER < 1800 */
140 #define strcasecmp(s1, s2) _stricmp(s1, s2)
141 #endif
142 /*@}*/
143
144
145 /*
146 * signbit() is a macro on Linux. Not available on Windows.
147 */
148 #ifndef signbit
149 #define signbit(x) ((x) < 0.0f)
150 #endif
151
152
153 /** single-precision inverse square root */
154 static inline float
155 INV_SQRTF(float x)
156 {
157 /* XXX we could try Quake's fast inverse square root function here */
158 return 1.0F / sqrtf(x);
159 }
160
161
162 /***
163 *** LOG2: Log base 2 of float
164 ***/
165 static inline GLfloat LOG2(GLfloat x)
166 {
167 #if 0
168 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
169 * Based on code from http://www.stereopsis.com/log2.html
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 #endif
177 /* Pretty fast, and accurate.
178 * Based on code from http://www.flipcode.com/totd/
179 */
180 fi_type num;
181 GLint log_2;
182 num.f = x;
183 log_2 = ((num.i >> 23) & 255) - 128;
184 num.i &= ~(255 << 23);
185 num.i += 127 << 23;
186 num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3;
187 return num.f + log_2;
188 }
189
190
191
192 /***
193 *** IS_INF_OR_NAN: test if float is infinite or NaN
194 ***/
195 #if defined(isfinite)
196 #define IS_INF_OR_NAN(x) (!isfinite(x))
197 #elif defined(finite)
198 #define IS_INF_OR_NAN(x) (!finite(x))
199 #elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
200 #define IS_INF_OR_NAN(x) (!isfinite(x))
201 #else
202 #define IS_INF_OR_NAN(x) (!finite(x))
203 #endif
204
205
206 /**
207 * Convert float to int by rounding to nearest integer, away from zero.
208 */
209 static inline int IROUND(float f)
210 {
211 return (int) ((f >= 0.0F) ? (f + 0.5F) : (f - 0.5F));
212 }
213
214
215 /**
216 * Convert float to int64 by rounding to nearest integer.
217 */
218 static inline GLint64 IROUND64(float f)
219 {
220 return (GLint64) ((f >= 0.0F) ? (f + 0.5F) : (f - 0.5F));
221 }
222
223
224 /**
225 * Convert positive float to int by rounding to nearest integer.
226 */
227 static inline int IROUND_POS(float f)
228 {
229 assert(f >= 0.0F);
230 return (int) (f + 0.5F);
231 }
232
233 #ifdef __x86_64__
234 # include <xmmintrin.h>
235 #endif
236
237 /**
238 * Convert float to int using a fast method. The rounding mode may vary.
239 */
240 static inline int F_TO_I(float f)
241 {
242 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
243 int r;
244 __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st");
245 return r;
246 #elif defined(USE_X86_ASM) && defined(_MSC_VER)
247 int r;
248 _asm {
249 fld f
250 fistp r
251 }
252 return r;
253 #elif defined(__x86_64__)
254 return _mm_cvt_ss2si(_mm_load_ss(&f));
255 #else
256 return IROUND(f);
257 #endif
258 }
259
260
261 /** Return (as an integer) floor of float */
262 static inline int IFLOOR(float f)
263 {
264 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
265 /*
266 * IEEE floor for computers that round to nearest or even.
267 * 'f' must be between -4194304 and 4194303.
268 * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
269 * but uses some IEEE specific tricks for better speed.
270 * Contributed by Josh Vanderhoof
271 */
272 int ai, bi;
273 double af, bf;
274 af = (3 << 22) + 0.5 + (double)f;
275 bf = (3 << 22) + 0.5 - (double)f;
276 /* GCC generates an extra fstp/fld without this. */
277 __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
278 __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
279 return (ai - bi) >> 1;
280 #else
281 int ai, bi;
282 double af, bf;
283 fi_type u;
284 af = (3 << 22) + 0.5 + (double)f;
285 bf = (3 << 22) + 0.5 - (double)f;
286 u.f = (float) af; ai = u.i;
287 u.f = (float) bf; bi = u.i;
288 return (ai - bi) >> 1;
289 #endif
290 }
291
292
293 /** Return (as an integer) ceiling of float */
294 static inline int ICEIL(float f)
295 {
296 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
297 /*
298 * IEEE ceil for computers that round to nearest or even.
299 * 'f' must be between -4194304 and 4194303.
300 * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
301 * but uses some IEEE specific tricks for better speed.
302 * Contributed by Josh Vanderhoof
303 */
304 int ai, bi;
305 double af, bf;
306 af = (3 << 22) + 0.5 + (double)f;
307 bf = (3 << 22) + 0.5 - (double)f;
308 /* GCC generates an extra fstp/fld without this. */
309 __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
310 __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
311 return (ai - bi + 1) >> 1;
312 #else
313 int ai, bi;
314 double af, bf;
315 fi_type u;
316 af = (3 << 22) + 0.5 + (double)f;
317 bf = (3 << 22) + 0.5 - (double)f;
318 u.f = (float) af; ai = u.i;
319 u.f = (float) bf; bi = u.i;
320 return (ai - bi + 1) >> 1;
321 #endif
322 }
323
324
325 /**
326 * Is x a power of two?
327 */
328 static inline int
329 _mesa_is_pow_two(int x)
330 {
331 return !(x & (x - 1));
332 }
333
334 /**
335 * Round given integer to next higer power of two
336 * If X is zero result is undefined.
337 *
338 * Source for the fallback implementation is
339 * Sean Eron Anderson's webpage "Bit Twiddling Hacks"
340 * http://graphics.stanford.edu/~seander/bithacks.html
341 *
342 * When using builtin function have to do some work
343 * for case when passed values 1 to prevent hiting
344 * undefined result from __builtin_clz. Undefined
345 * results would be different depending on optimization
346 * level used for build.
347 */
348 static inline int32_t
349 _mesa_next_pow_two_32(uint32_t x)
350 {
351 #ifdef HAVE___BUILTIN_CLZ
352 uint32_t y = (x != 1);
353 return (1 + y) << ((__builtin_clz(x - y) ^ 31) );
354 #else
355 x--;
356 x |= x >> 1;
357 x |= x >> 2;
358 x |= x >> 4;
359 x |= x >> 8;
360 x |= x >> 16;
361 x++;
362 return x;
363 #endif
364 }
365
366 static inline int64_t
367 _mesa_next_pow_two_64(uint64_t x)
368 {
369 #ifdef HAVE___BUILTIN_CLZLL
370 uint64_t y = (x != 1);
371 STATIC_ASSERT(sizeof(x) == sizeof(long long));
372 return (1 + y) << ((__builtin_clzll(x - y) ^ 63));
373 #else
374 x--;
375 x |= x >> 1;
376 x |= x >> 2;
377 x |= x >> 4;
378 x |= x >> 8;
379 x |= x >> 16;
380 x |= x >> 32;
381 x++;
382 return x;
383 #endif
384 }
385
386
387 /*
388 * Returns the floor form of binary logarithm for a 32-bit integer.
389 */
390 static inline GLuint
391 _mesa_logbase2(GLuint n)
392 {
393 #ifdef HAVE___BUILTIN_CLZ
394 return (31 - __builtin_clz(n | 1));
395 #else
396 GLuint pos = 0;
397 if (n >= 1<<16) { n >>= 16; pos += 16; }
398 if (n >= 1<< 8) { n >>= 8; pos += 8; }
399 if (n >= 1<< 4) { n >>= 4; pos += 4; }
400 if (n >= 1<< 2) { n >>= 2; pos += 2; }
401 if (n >= 1<< 1) { pos += 1; }
402 return pos;
403 #endif
404 }
405
406
407 /**
408 * Return 1 if this is a little endian machine, 0 if big endian.
409 */
410 static inline GLboolean
411 _mesa_little_endian(void)
412 {
413 const GLuint ui = 1; /* intentionally not static */
414 return *((const GLubyte *) &ui);
415 }
416
417
418
419 /**********************************************************************
420 * Functions
421 */
422
423 extern void *
424 _mesa_align_malloc( size_t bytes, unsigned long alignment );
425
426 extern void *
427 _mesa_align_calloc( size_t bytes, unsigned long alignment );
428
429 extern void
430 _mesa_align_free( void *ptr );
431
432 extern void *
433 _mesa_align_realloc(void *oldBuffer, size_t oldSize, size_t newSize,
434 unsigned long alignment);
435
436 extern void *
437 _mesa_exec_malloc( GLuint size );
438
439 extern void
440 _mesa_exec_free( void *addr );
441
442
443 #ifndef FFS_DEFINED
444 #define FFS_DEFINED 1
445 #ifdef HAVE___BUILTIN_FFS
446 #define ffs __builtin_ffs
447 #else
448 extern int ffs(int i);
449 #endif
450
451 #ifdef HAVE___BUILTIN_FFSLL
452 #define ffsll __builtin_ffsll
453 #else
454 extern int ffsll(long long int i);
455 #endif
456 #endif /* FFS_DEFINED */
457
458
459 #ifdef HAVE___BUILTIN_POPCOUNT
460 #define _mesa_bitcount(i) __builtin_popcount(i)
461 #else
462 extern unsigned int
463 _mesa_bitcount(unsigned int n);
464 #endif
465
466 #ifdef HAVE___BUILTIN_POPCOUNTLL
467 #define _mesa_bitcount_64(i) __builtin_popcountll(i)
468 #else
469 extern unsigned int
470 _mesa_bitcount_64(uint64_t n);
471 #endif
472
473 /**
474 * Find the last (most significant) bit set in a word.
475 *
476 * Essentially ffs() in the reverse direction.
477 */
478 static inline unsigned int
479 _mesa_fls(unsigned int n)
480 {
481 #ifdef HAVE___BUILTIN_CLZ
482 return n == 0 ? 0 : 32 - __builtin_clz(n);
483 #else
484 unsigned int v = 1;
485
486 if (n == 0)
487 return 0;
488
489 while (n >>= 1)
490 v++;
491
492 return v;
493 #endif
494 }
495
496 extern int
497 _mesa_round_to_even(float val);
498
499 extern GLhalfARB
500 _mesa_float_to_half(float f);
501
502 extern float
503 _mesa_half_to_float(GLhalfARB h);
504
505 static inline bool
506 _mesa_half_is_negative(GLhalfARB h)
507 {
508 return h & 0x8000;
509 }
510
511 extern char *
512 _mesa_strdup( const char *s );
513
514 extern unsigned int
515 _mesa_str_checksum(const char *str);
516
517 extern int
518 _mesa_snprintf( char *str, size_t size, const char *fmt, ... ) PRINTFLIKE(3, 4);
519
520 extern int
521 _mesa_vsnprintf(char *str, size_t size, const char *fmt, va_list arg);
522
523
524 #if defined(_MSC_VER) && !defined(snprintf)
525 #define snprintf _snprintf
526 #endif
527
528
529 #ifdef __cplusplus
530 }
531 #endif
532
533
534 #endif /* IMPORTS_H */