mesa: Replace a priori knowledge of gcc builtins with configure tests.
[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 *** CEILF: ceiling of float
208 *** FLOORF: floor of float
209 *** FABSF: absolute value of float
210 *** LOGF: the natural logarithm (base e) of the value
211 *** EXPF: raise e to the value
212 *** LDEXPF: multiply value by an integral power of two
213 *** FREXPF: extract mantissa and exponent from value
214 ***/
215 #if defined(__gnu_linux__)
216 /* C99 functions */
217 #define CEILF(x) ceilf(x)
218 #define FLOORF(x) floorf(x)
219 #define FABSF(x) fabsf(x)
220 #define LOGF(x) logf(x)
221 #define EXPF(x) expf(x)
222 #define LDEXPF(x,y) ldexpf(x,y)
223 #define FREXPF(x,y) frexpf(x,y)
224 #else
225 #define CEILF(x) ((GLfloat) ceil(x))
226 #define FLOORF(x) ((GLfloat) floor(x))
227 #define FABSF(x) ((GLfloat) fabs(x))
228 #define LOGF(x) ((GLfloat) log(x))
229 #define EXPF(x) ((GLfloat) exp(x))
230 #define LDEXPF(x,y) ((GLfloat) ldexp(x,y))
231 #define FREXPF(x,y) ((GLfloat) frexp(x,y))
232 #endif
233
234
235 /**
236 * Convert float to int by rounding to nearest integer, away from zero.
237 */
238 static inline int IROUND(float f)
239 {
240 return (int) ((f >= 0.0F) ? (f + 0.5F) : (f - 0.5F));
241 }
242
243
244 /**
245 * Convert float to int64 by rounding to nearest integer.
246 */
247 static inline GLint64 IROUND64(float f)
248 {
249 return (GLint64) ((f >= 0.0F) ? (f + 0.5F) : (f - 0.5F));
250 }
251
252
253 /**
254 * Convert positive float to int by rounding to nearest integer.
255 */
256 static inline int IROUND_POS(float f)
257 {
258 assert(f >= 0.0F);
259 return (int) (f + 0.5F);
260 }
261
262 #ifdef __x86_64__
263 # include <xmmintrin.h>
264 #endif
265
266 /**
267 * Convert float to int using a fast method. The rounding mode may vary.
268 */
269 static inline int F_TO_I(float f)
270 {
271 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
272 int r;
273 __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st");
274 return r;
275 #elif defined(USE_X86_ASM) && defined(_MSC_VER)
276 int r;
277 _asm {
278 fld f
279 fistp r
280 }
281 return r;
282 #elif defined(__x86_64__)
283 return _mm_cvt_ss2si(_mm_load_ss(&f));
284 #else
285 return IROUND(f);
286 #endif
287 }
288
289
290 /** Return (as an integer) floor of float */
291 static inline int IFLOOR(float f)
292 {
293 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
294 /*
295 * IEEE floor for computers that round to nearest or even.
296 * 'f' must be between -4194304 and 4194303.
297 * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
298 * but uses some IEEE specific tricks for better speed.
299 * Contributed by Josh Vanderhoof
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 #else
310 int ai, bi;
311 double af, bf;
312 fi_type u;
313 af = (3 << 22) + 0.5 + (double)f;
314 bf = (3 << 22) + 0.5 - (double)f;
315 u.f = (float) af; ai = u.i;
316 u.f = (float) bf; bi = u.i;
317 return (ai - bi) >> 1;
318 #endif
319 }
320
321
322 /** Return (as an integer) ceiling of float */
323 static inline int ICEIL(float f)
324 {
325 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
326 /*
327 * IEEE ceil for computers that round to nearest or even.
328 * 'f' must be between -4194304 and 4194303.
329 * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
330 * but uses some IEEE specific tricks for better speed.
331 * Contributed by Josh Vanderhoof
332 */
333 int ai, bi;
334 double af, bf;
335 af = (3 << 22) + 0.5 + (double)f;
336 bf = (3 << 22) + 0.5 - (double)f;
337 /* GCC generates an extra fstp/fld without this. */
338 __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
339 __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
340 return (ai - bi + 1) >> 1;
341 #else
342 int ai, bi;
343 double af, bf;
344 fi_type u;
345 af = (3 << 22) + 0.5 + (double)f;
346 bf = (3 << 22) + 0.5 - (double)f;
347 u.f = (float) af; ai = u.i;
348 u.f = (float) bf; bi = u.i;
349 return (ai - bi + 1) >> 1;
350 #endif
351 }
352
353
354 /**
355 * Is x a power of two?
356 */
357 static inline int
358 _mesa_is_pow_two(int x)
359 {
360 return !(x & (x - 1));
361 }
362
363 /**
364 * Round given integer to next higer power of two
365 * If X is zero result is undefined.
366 *
367 * Source for the fallback implementation is
368 * Sean Eron Anderson's webpage "Bit Twiddling Hacks"
369 * http://graphics.stanford.edu/~seander/bithacks.html
370 *
371 * When using builtin function have to do some work
372 * for case when passed values 1 to prevent hiting
373 * undefined result from __builtin_clz. Undefined
374 * results would be different depending on optimization
375 * level used for build.
376 */
377 static inline int32_t
378 _mesa_next_pow_two_32(uint32_t x)
379 {
380 #ifdef HAVE___BUILTIN_CLZ
381 uint32_t y = (x != 1);
382 return (1 + y) << ((__builtin_clz(x - y) ^ 31) );
383 #else
384 x--;
385 x |= x >> 1;
386 x |= x >> 2;
387 x |= x >> 4;
388 x |= x >> 8;
389 x |= x >> 16;
390 x++;
391 return x;
392 #endif
393 }
394
395 static inline int64_t
396 _mesa_next_pow_two_64(uint64_t x)
397 {
398 #ifdef HAVE___BUILTIN_CLZLL
399 uint64_t y = (x != 1);
400 STATIC_ASSERT(sizeof(x) == sizeof(long long));
401 return (1 + y) << ((__builtin_clzll(x - y) ^ 63));
402 #else
403 x--;
404 x |= x >> 1;
405 x |= x >> 2;
406 x |= x >> 4;
407 x |= x >> 8;
408 x |= x >> 16;
409 x |= x >> 32;
410 x++;
411 return x;
412 #endif
413 }
414
415
416 /*
417 * Returns the floor form of binary logarithm for a 32-bit integer.
418 */
419 static inline GLuint
420 _mesa_logbase2(GLuint n)
421 {
422 #ifdef HAVE___BUILTIN_CLZ
423 return (31 - __builtin_clz(n | 1));
424 #else
425 GLuint pos = 0;
426 if (n >= 1<<16) { n >>= 16; pos += 16; }
427 if (n >= 1<< 8) { n >>= 8; pos += 8; }
428 if (n >= 1<< 4) { n >>= 4; pos += 4; }
429 if (n >= 1<< 2) { n >>= 2; pos += 2; }
430 if (n >= 1<< 1) { pos += 1; }
431 return pos;
432 #endif
433 }
434
435
436 /**
437 * Return 1 if this is a little endian machine, 0 if big endian.
438 */
439 static inline GLboolean
440 _mesa_little_endian(void)
441 {
442 const GLuint ui = 1; /* intentionally not static */
443 return *((const GLubyte *) &ui);
444 }
445
446
447
448 /**********************************************************************
449 * Functions
450 */
451
452 extern void *
453 _mesa_align_malloc( size_t bytes, unsigned long alignment );
454
455 extern void *
456 _mesa_align_calloc( size_t bytes, unsigned long alignment );
457
458 extern void
459 _mesa_align_free( void *ptr );
460
461 extern void *
462 _mesa_align_realloc(void *oldBuffer, size_t oldSize, size_t newSize,
463 unsigned long alignment);
464
465 extern void *
466 _mesa_exec_malloc( GLuint size );
467
468 extern void
469 _mesa_exec_free( void *addr );
470
471
472 #ifndef FFS_DEFINED
473 #define FFS_DEFINED 1
474 #ifdef HAVE___BUILTIN_FFS
475 #define ffs __builtin_ffs
476 #else
477 extern int ffs(int i);
478 #endif
479
480 #ifdef HAVE___BUILTIN_FFSLL
481 #define ffsll __builtin_ffsll
482 #else
483 extern int ffsll(long long int i);
484 #endif
485 #endif /* FFS_DEFINED */
486
487
488 #ifdef HAVE___BUILTIN_POPCOUNT
489 #define _mesa_bitcount(i) __builtin_popcount(i)
490 #else
491 extern unsigned int
492 _mesa_bitcount(unsigned int n);
493 #endif
494
495 #ifdef HAVE___BUILTIN_POPCOUNTLL
496 #define _mesa_bitcount_64(i) __builtin_popcountll(i)
497 #else
498 extern unsigned int
499 _mesa_bitcount_64(uint64_t n);
500 #endif
501
502 /**
503 * Find the last (most significant) bit set in a word.
504 *
505 * Essentially ffs() in the reverse direction.
506 */
507 static inline unsigned int
508 _mesa_fls(unsigned int n)
509 {
510 #ifdef HAVE___BUILTIN_CLZ
511 return n == 0 ? 0 : 32 - __builtin_clz(n);
512 #else
513 unsigned int v = 1;
514
515 if (n == 0)
516 return 0;
517
518 while (n >>= 1)
519 v++;
520
521 return v;
522 #endif
523 }
524
525 extern int
526 _mesa_round_to_even(float val);
527
528 extern GLhalfARB
529 _mesa_float_to_half(float f);
530
531 extern float
532 _mesa_half_to_float(GLhalfARB h);
533
534 static inline bool
535 _mesa_half_is_negative(GLhalfARB h)
536 {
537 return h & 0x8000;
538 }
539
540 extern char *
541 _mesa_strdup( const char *s );
542
543 extern float
544 _mesa_strtof( const char *s, char **end );
545
546 extern unsigned int
547 _mesa_str_checksum(const char *str);
548
549 extern int
550 _mesa_snprintf( char *str, size_t size, const char *fmt, ... ) PRINTFLIKE(3, 4);
551
552 extern int
553 _mesa_vsnprintf(char *str, size_t size, const char *fmt, va_list arg);
554
555
556 #if defined(_MSC_VER) && !defined(snprintf)
557 #define snprintf _snprintf
558 #endif
559
560
561 #ifdef __cplusplus
562 }
563 #endif
564
565
566 #endif /* IMPORTS_H */