mesa: remove platform checks around __builtin_ffs, __builtin_ffsll
[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 #include "errors.h"
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; 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 * \name Work-arounds for platforms that lack C99 math functions
101 */
102 /*@{*/
103 #if (!defined(_XOPEN_SOURCE) || (_XOPEN_SOURCE < 600)) && !defined(_ISOC99_SOURCE) \
104 && (!defined(__STDC_VERSION__) || (__STDC_VERSION__ < 199901L)) \
105 && (!defined(_MSC_VER) || (_MSC_VER < 1400))
106 #define acosf(f) ((float) acos(f))
107 #define asinf(f) ((float) asin(f))
108 #define atan2f(x,y) ((float) atan2(x,y))
109 #define atanf(f) ((float) atan(f))
110 #define ceilf(f) ((float) ceil(f))
111 #define cosf(f) ((float) cos(f))
112 #define coshf(f) ((float) cosh(f))
113 #define expf(f) ((float) exp(f))
114 #define exp2f(f) ((float) exp2(f))
115 #define floorf(f) ((float) floor(f))
116 #define logf(f) ((float) log(f))
117
118 #ifdef ANDROID
119 #define log2f(f) (logf(f) * (float) (1.0 / M_LN2))
120 #else
121 #define log2f(f) ((float) log2(f))
122 #endif
123
124 #define powf(x,y) ((float) pow(x,y))
125 #define sinf(f) ((float) sin(f))
126 #define sinhf(f) ((float) sinh(f))
127 #define sqrtf(f) ((float) sqrt(f))
128 #define tanf(f) ((float) tan(f))
129 #define tanhf(f) ((float) tanh(f))
130 #define acoshf(f) ((float) acosh(f))
131 #define asinhf(f) ((float) asinh(f))
132 #define atanhf(f) ((float) atanh(f))
133 #endif
134
135 #if defined(_MSC_VER)
136 static inline float truncf(float x) { return x < 0.0f ? ceilf(x) : floorf(x); }
137 static inline float exp2f(float x) { return powf(2.0f, x); }
138 static inline float log2f(float x) { return logf(x) * 1.442695041f; }
139 static inline float asinhf(float x) { return logf(x + sqrtf(x * x + 1.0f)); }
140 static inline float acoshf(float x) { return logf(x + sqrtf(x * x - 1.0f)); }
141 static inline float atanhf(float x) { return (logf(1.0f + x) - logf(1.0f - x)) / 2.0f; }
142 static inline int isblank(int ch) { return ch == ' ' || ch == '\t'; }
143 #define strtoll(p, e, b) _strtoi64(p, e, b)
144 #endif
145 /*@}*/
146
147
148 /*
149 * signbit() is a macro on Linux. Not available on Windows.
150 */
151 #ifndef signbit
152 #define signbit(x) ((x) < 0.0f)
153 #endif
154
155
156 /** single-precision inverse square root */
157 static inline float
158 INV_SQRTF(float x)
159 {
160 /* XXX we could try Quake's fast inverse square root function here */
161 return 1.0F / sqrtf(x);
162 }
163
164
165 /***
166 *** LOG2: Log base 2 of float
167 ***/
168 static inline GLfloat LOG2(GLfloat x)
169 {
170 #ifdef USE_IEEE
171 #if 0
172 /* This is pretty fast, but not accurate enough (only 2 fractional bits).
173 * Based on code from http://www.stereopsis.com/log2.html
174 */
175 const GLfloat y = x * x * x * x;
176 const GLuint ix = *((GLuint *) &y);
177 const GLuint exp = (ix >> 23) & 0xFF;
178 const GLint log2 = ((GLint) exp) - 127;
179 return (GLfloat) log2 * (1.0 / 4.0); /* 4, because of x^4 above */
180 #endif
181 /* Pretty fast, and accurate.
182 * Based on code from http://www.flipcode.com/totd/
183 */
184 fi_type num;
185 GLint log_2;
186 num.f = x;
187 log_2 = ((num.i >> 23) & 255) - 128;
188 num.i &= ~(255 << 23);
189 num.i += 127 << 23;
190 num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3;
191 return num.f + log_2;
192 #else
193 /*
194 * NOTE: log_base_2(x) = log(x) / log(2)
195 * NOTE: 1.442695 = 1/log(2).
196 */
197 return (GLfloat) (log(x) * 1.442695F);
198 #endif
199 }
200
201
202
203 /***
204 *** IS_INF_OR_NAN: test if float is infinite or NaN
205 ***/
206 #ifdef USE_IEEE
207 static inline int IS_INF_OR_NAN( float x )
208 {
209 fi_type tmp;
210 tmp.f = x;
211 return !(int)((unsigned int)((tmp.i & 0x7fffffff)-0x7f800000) >> 31);
212 }
213 #elif defined(isfinite)
214 #define IS_INF_OR_NAN(x) (!isfinite(x))
215 #elif defined(finite)
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 *** CEILF: ceiling of float
226 *** FLOORF: floor of float
227 *** FABSF: absolute value of float
228 *** LOGF: the natural logarithm (base e) of the value
229 *** EXPF: raise e to the value
230 *** LDEXPF: multiply value by an integral power of two
231 *** FREXPF: extract mantissa and exponent from value
232 ***/
233 #if defined(__gnu_linux__)
234 /* C99 functions */
235 #define CEILF(x) ceilf(x)
236 #define FLOORF(x) floorf(x)
237 #define FABSF(x) fabsf(x)
238 #define LOGF(x) logf(x)
239 #define EXPF(x) expf(x)
240 #define LDEXPF(x,y) ldexpf(x,y)
241 #define FREXPF(x,y) frexpf(x,y)
242 #else
243 #define CEILF(x) ((GLfloat) ceil(x))
244 #define FLOORF(x) ((GLfloat) floor(x))
245 #define FABSF(x) ((GLfloat) fabs(x))
246 #define LOGF(x) ((GLfloat) log(x))
247 #define EXPF(x) ((GLfloat) exp(x))
248 #define LDEXPF(x,y) ((GLfloat) ldexp(x,y))
249 #define FREXPF(x,y) ((GLfloat) frexp(x,y))
250 #endif
251
252
253 /**
254 * Convert float to int by rounding to nearest integer, away from zero.
255 */
256 static inline int IROUND(float f)
257 {
258 return (int) ((f >= 0.0F) ? (f + 0.5F) : (f - 0.5F));
259 }
260
261
262 /**
263 * Convert float to int64 by rounding to nearest integer.
264 */
265 static inline GLint64 IROUND64(float f)
266 {
267 return (GLint64) ((f >= 0.0F) ? (f + 0.5F) : (f - 0.5F));
268 }
269
270
271 /**
272 * Convert positive float to int by rounding to nearest integer.
273 */
274 static inline int IROUND_POS(float f)
275 {
276 assert(f >= 0.0F);
277 return (int) (f + 0.5F);
278 }
279
280
281 /**
282 * Convert float to int using a fast method. The rounding mode may vary.
283 * XXX We could use an x86-64/SSE2 version here.
284 */
285 static inline int F_TO_I(float f)
286 {
287 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
288 int r;
289 __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st");
290 return r;
291 #elif defined(USE_X86_ASM) && defined(_MSC_VER)
292 int r;
293 _asm {
294 fld f
295 fistp r
296 }
297 return r;
298 #else
299 return IROUND(f);
300 #endif
301 }
302
303
304 /** Return (as an integer) floor of float */
305 static inline int IFLOOR(float f)
306 {
307 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
308 /*
309 * IEEE floor for computers that round to nearest or even.
310 * 'f' must be between -4194304 and 4194303.
311 * This floor operation is done by "(iround(f + .5) + iround(f - .5)) >> 1",
312 * but uses some IEEE specific tricks for better speed.
313 * Contributed by Josh Vanderhoof
314 */
315 int ai, bi;
316 double af, bf;
317 af = (3 << 22) + 0.5 + (double)f;
318 bf = (3 << 22) + 0.5 - (double)f;
319 /* GCC generates an extra fstp/fld without this. */
320 __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
321 __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
322 return (ai - bi) >> 1;
323 #elif defined(USE_IEEE)
324 int ai, bi;
325 double af, bf;
326 fi_type u;
327 af = (3 << 22) + 0.5 + (double)f;
328 bf = (3 << 22) + 0.5 - (double)f;
329 u.f = (float) af; ai = u.i;
330 u.f = (float) bf; bi = u.i;
331 return (ai - bi) >> 1;
332 #else
333 int i = IROUND(f);
334 return (i > f) ? i - 1 : i;
335 #endif
336 }
337
338
339 /** Return (as an integer) ceiling of float */
340 static inline int ICEIL(float f)
341 {
342 #if defined(USE_X86_ASM) && defined(__GNUC__) && defined(__i386__)
343 /*
344 * IEEE ceil for computers that round to nearest or even.
345 * 'f' must be between -4194304 and 4194303.
346 * This ceil operation is done by "(iround(f + .5) + iround(f - .5) + 1) >> 1",
347 * but uses some IEEE specific tricks for better speed.
348 * Contributed by Josh Vanderhoof
349 */
350 int ai, bi;
351 double af, bf;
352 af = (3 << 22) + 0.5 + (double)f;
353 bf = (3 << 22) + 0.5 - (double)f;
354 /* GCC generates an extra fstp/fld without this. */
355 __asm__ ("fstps %0" : "=m" (ai) : "t" (af) : "st");
356 __asm__ ("fstps %0" : "=m" (bi) : "t" (bf) : "st");
357 return (ai - bi + 1) >> 1;
358 #elif defined(USE_IEEE)
359 int ai, bi;
360 double af, bf;
361 fi_type u;
362 af = (3 << 22) + 0.5 + (double)f;
363 bf = (3 << 22) + 0.5 - (double)f;
364 u.f = (float) af; ai = u.i;
365 u.f = (float) bf; bi = u.i;
366 return (ai - bi + 1) >> 1;
367 #else
368 int i = IROUND(f);
369 return (i < f) ? i + 1 : i;
370 #endif
371 }
372
373
374 /**
375 * Is x a power of two?
376 */
377 static inline int
378 _mesa_is_pow_two(int x)
379 {
380 return !(x & (x - 1));
381 }
382
383 /**
384 * Round given integer to next higer power of two
385 * If X is zero result is undefined.
386 *
387 * Source for the fallback implementation is
388 * Sean Eron Anderson's webpage "Bit Twiddling Hacks"
389 * http://graphics.stanford.edu/~seander/bithacks.html
390 *
391 * When using builtin function have to do some work
392 * for case when passed values 1 to prevent hiting
393 * undefined result from __builtin_clz. Undefined
394 * results would be different depending on optimization
395 * level used for build.
396 */
397 static inline int32_t
398 _mesa_next_pow_two_32(uint32_t x)
399 {
400 #if defined(__GNUC__) && \
401 ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
402 uint32_t y = (x != 1);
403 return (1 + y) << ((__builtin_clz(x - y) ^ 31) );
404 #else
405 x--;
406 x |= x >> 1;
407 x |= x >> 2;
408 x |= x >> 4;
409 x |= x >> 8;
410 x |= x >> 16;
411 x++;
412 return x;
413 #endif
414 }
415
416 static inline int64_t
417 _mesa_next_pow_two_64(uint64_t x)
418 {
419 #if defined(__GNUC__) && \
420 ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
421 uint64_t y = (x != 1);
422 if (sizeof(x) == sizeof(long))
423 return (1 + y) << ((__builtin_clzl(x - y) ^ 63));
424 else
425 return (1 + y) << ((__builtin_clzll(x - y) ^ 63));
426 #else
427 x--;
428 x |= x >> 1;
429 x |= x >> 2;
430 x |= x >> 4;
431 x |= x >> 8;
432 x |= x >> 16;
433 x |= x >> 32;
434 x++;
435 return x;
436 #endif
437 }
438
439
440 /*
441 * Returns the floor form of binary logarithm for a 32-bit integer.
442 */
443 static inline GLuint
444 _mesa_logbase2(GLuint n)
445 {
446 #if defined(__GNUC__) && \
447 ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
448 return (31 - __builtin_clz(n | 1));
449 #else
450 GLuint pos = 0;
451 if (n >= 1<<16) { n >>= 16; pos += 16; }
452 if (n >= 1<< 8) { n >>= 8; pos += 8; }
453 if (n >= 1<< 4) { n >>= 4; pos += 4; }
454 if (n >= 1<< 2) { n >>= 2; pos += 2; }
455 if (n >= 1<< 1) { pos += 1; }
456 return pos;
457 #endif
458 }
459
460
461 /**
462 * Return 1 if this is a little endian machine, 0 if big endian.
463 */
464 static inline GLboolean
465 _mesa_little_endian(void)
466 {
467 const GLuint ui = 1; /* intentionally not static */
468 return *((const GLubyte *) &ui);
469 }
470
471
472
473 /**********************************************************************
474 * Functions
475 */
476
477 extern void *
478 _mesa_align_malloc( size_t bytes, unsigned long alignment );
479
480 extern void *
481 _mesa_align_calloc( size_t bytes, unsigned long alignment );
482
483 extern void
484 _mesa_align_free( void *ptr );
485
486 extern void *
487 _mesa_align_realloc(void *oldBuffer, size_t oldSize, size_t newSize,
488 unsigned long alignment);
489
490 extern void *
491 _mesa_exec_malloc( GLuint size );
492
493 extern void
494 _mesa_exec_free( void *addr );
495
496 extern void *
497 _mesa_realloc( void *oldBuffer, size_t oldSize, size_t newSize );
498
499
500 #ifndef FFS_DEFINED
501 #define FFS_DEFINED 1
502 #ifdef __GNUC__
503 #define ffs __builtin_ffs
504 #define ffsll __builtin_ffsll
505 #else
506 extern int ffs(int i);
507 extern int ffsll(long long int i);
508 #endif /*__ GNUC__ */
509 #endif /* FFS_DEFINED */
510
511
512 #if defined(__GNUC__) && ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
513 #define _mesa_bitcount(i) __builtin_popcount(i)
514 #define _mesa_bitcount_64(i) __builtin_popcountll(i)
515 #else
516 extern unsigned int
517 _mesa_bitcount(unsigned int n);
518 extern unsigned int
519 _mesa_bitcount_64(uint64_t n);
520 #endif
521
522 /**
523 * Find the last (most significant) bit set in a word.
524 *
525 * Essentially ffs() in the reverse direction.
526 */
527 static inline unsigned int
528 _mesa_fls(unsigned int n)
529 {
530 #if defined(__GNUC__) && ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304)
531 return n == 0 ? 0 : 32 - __builtin_clz(n);
532 #else
533 unsigned int v = 1;
534
535 if (n == 0)
536 return 0;
537
538 while (n >>= 1)
539 v++;
540
541 return v;
542 #endif
543 }
544
545 extern int
546 _mesa_round_to_even(float val);
547
548 extern GLhalfARB
549 _mesa_float_to_half(float f);
550
551 extern float
552 _mesa_half_to_float(GLhalfARB h);
553
554
555 extern void *
556 _mesa_bsearch( const void *key, const void *base, size_t nmemb, size_t size,
557 int (*compar)(const void *, const void *) );
558
559 extern char *
560 _mesa_getenv( const char *var );
561
562 extern char *
563 _mesa_strdup( const char *s );
564
565 extern float
566 _mesa_strtof( const char *s, char **end );
567
568 extern unsigned int
569 _mesa_str_checksum(const char *str);
570
571 extern int
572 _mesa_snprintf( char *str, size_t size, const char *fmt, ... ) PRINTFLIKE(3, 4);
573
574 extern int
575 _mesa_vsnprintf(char *str, size_t size, const char *fmt, va_list arg);
576
577
578 #if defined(_MSC_VER) && !defined(snprintf)
579 #define snprintf _snprintf
580 #endif
581
582
583 #ifdef __cplusplus
584 }
585 #endif
586
587
588 #endif /* IMPORTS_H */